SPB Git
spboucher.ai

spb/wp33_uqo

Public MIT

Transport: native-dtype activations (bfloat16 exact for Qwen3); LOG Gate F v1 discarded + stream fix

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Simon-Pierre Boucher committed (Nov 1, 2026) parent 9b97c26 Browse files

13 changed files +1,098 −15

13 changed files
Unified Split
added benchmarks/raw/gateF/traces_qwen3_m3u96a/index.json +631 −0
@@ -0,0 +1,631 @@
1 +{
2 + "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa",
3 + "L": 48,
4 + "E": 128,
5 + "K": 8,
6 + "max_tokens": 128,
7 + "ts": "2026-11-01T01:18:01",
8 + "cmd": "benchmarks/scripts/collect_traces.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workloads benchmarks/configs/workloads_v2.yaml --max-tokens 128 --out benchmarks/raw/gateF/traces_qwen3_m3u96a",
9 + "prompts": [
10 + {
11 + "name": "general_fit_0",
12 + "family": "general",
13 + "split": "fit",
14 + "prompt": "Explain in three sentences why mixture-of-experts models are memory-bound during",
15 + "n_prompt": 17,
16 + "n_gen": 128,
17 + "seconds": 2.558898749994114,
18 + "text": " Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on a large number of expert subnetwo"
19 + },
20 + {
21 + "name": "general_fit_1",
22 + "family": "general",
23 + "split": "fit",
24 + "prompt": "Write a short paragraph describing the climate of Gatineau, Qu\u00e9bec, across the f",
25 + "n_prompt": 19,
26 + "n_gen": 128,
27 + "seconds": 2.3968757909024134,
28 + "text": " Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continen"
29 + },
30 + {
31 + "name": "general_fit_2",
32 + "family": "general",
33 + "split": "fit",
34 + "prompt": "Give five practical tips for a first-year university student learning to manage ",
35 + "n_prompt": 16,
36 + "n_gen": 128,
37 + "seconds": 2.398546750075184,
38 + "text": " 1. **Create a Weekly Schedule Using a Planner or Digital Calendar**: Start by mapping out your classes, study sessions,"
39 + },
40 + {
41 + "name": "general_fit_3",
42 + "family": "general",
43 + "split": "fit",
44 + "prompt": "Describe how a hydroelectric dam converts the energy of falling water into elect",
45 + "n_prompt": 15,
46 + "n_gen": 128,
47 + "seconds": 2.3993994999909773,
48 + "text": " Include the roles of the intake, turbine, generator, and transformer.\n\nIf the question has a final answer, please place"
49 + },
50 + {
51 + "name": "general_fit_4",
52 + "family": "general",
53 + "split": "fit",
54 + "prompt": "Write a friendly email inviting colleagues to a Friday afternoon seminar on hous",
55 + "n_prompt": 15,
56 + "n_gen": 128,
57 + "seconds": 2.386043583974242,
58 + "text": " Include a light-hearted note about coffee and a suggestion to bring a favorite snack.\n\nSubject: \ud83c\udf3f Join Us This Friday f"
59 + },
60 + {
61 + "name": "general_fit_5",
62 + "family": "general",
63 + "split": "fit",
64 + "prompt": "What are the main arguments for and against remote work in knowledge industries?",
65 + "n_prompt": 14,
66 + "n_gen": 128,
67 + "seconds": 2.400736250099726,
68 + "text": " Additionally, what are the potential benefits and drawbacks of remote work for employees, and how might remote work imp"
69 + },
70 + {
71 + "name": "general_fit_6",
72 + "family": "general",
73 + "split": "fit",
74 + "prompt": "Explain the difference between weather and climate to a ten-year-old.",
75 + "n_prompt": 14,
76 + "n_gen": 128,
77 + "seconds": 2.409935374977067,
78 + "text": " Sure! Let\u2019s imagine you\u2019re looking at a big picture of the sky.\n\n**Weather** is what\u2019s happening *right now* outside. L"
79 + },
80 + {
81 + "name": "general_fit_7",
82 + "family": "general",
83 + "split": "fit",
84 + "prompt": "Draft a two-paragraph product description for a quiet mechanical keyboard.",
85 + "n_prompt": 13,
86 + "n_gen": 128,
87 + "seconds": 2.3774168750969693,
88 + "text": " Begin the first paragraph with the phrase \"Engineered for precision and comfort,\" and end the second paragraph with the"
89 + },
90 + {
91 + "name": "general_fit_8",
92 + "family": "general",
93 + "split": "fit",
94 + "prompt": "List the steps to prepare a simple vegetable soup, with approximate timings.",
95 + "n_prompt": 14,
96 + "n_gen": 128,
97 + "seconds": 2.3827790000941604,
98 + "text": " Here\u2019s a simple and delicious vegetable soup recipe with approximate timings for each step:\n\n---\n\n**Simple Vegetable So"
99 + },
100 + {
101 + "name": "general_fit_9",
102 + "family": "general",
103 + "split": "fit",
104 + "prompt": "Why do leaves change colour in autumn? Answer in one paragraph.",
105 + "n_prompt": 13,
106 + "n_gen": 128,
107 + "seconds": 2.3947936670156196,
108 + "text": " Leaves change colour in autumn due to a combination of biological and environmental factors. As daylight hours shorten "
109 + },
110 + {
111 + "name": "general_heldout_0",
112 + "family": "general",
113 + "split": "heldout",
114 + "prompt": "Describe the main differences between a Mac Studio and a MacBook Pro for scienti",
115 + "n_prompt": 16,
116 + "n_gen": 128,
117 + "seconds": 2.4038165831007063,
118 + "text": " The Mac Studio and MacBook Pro are both high-performance Apple devices designed for demanding workloads, including scie"
119 + },
120 + {
121 + "name": "general_heldout_1",
122 + "family": "general",
123 + "split": "heldout",
124 + "prompt": "Summarize the history of the Rideau Canal in one paragraph.",
125 + "n_prompt": 14,
126 + "n_gen": 128,
127 + "seconds": 2.361865375074558,
128 + "text": " The Rideau Canal, located in Ontario, Canada, is a historic waterway that was constructed in the early 19th century as "
129 + },
130 + {
131 + "name": "general_heldout_2",
132 + "family": "general",
133 + "split": "heldout",
134 + "prompt": "Write a short motivational note for someone starting to learn the piano as an ad",
135 + "n_prompt": 16,
136 + "n_gen": 128,
137 + "seconds": 2.3701815000968054,
138 + "text": " Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nStarting to learn the"
139 + },
140 + {
141 + "name": "general_heldout_3",
142 + "family": "general",
143 + "split": "heldout",
144 + "prompt": "What should a visitor know before hiking in Gatineau Park in winter?",
145 + "n_prompt": 15,
146 + "n_gen": 128,
147 + "seconds": 2.4134104170370847,
148 + "text": " What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions?"
149 + },
150 + {
151 + "name": "code_fit_0",
152 + "family": "code",
153 + "split": "fit",
154 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"\"\n",
155 + "n_prompt": 13,
156 + "n_gen": 118,
157 + "seconds": 2.211064000031911,
158 + "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n "
159 + },
160 + {
161 + "name": "code_fit_1",
162 + "family": "code",
163 + "split": "fit",
164 + "prompt": "Write a Python function that parses a CSV file of timestamps and returns the med",
165 + "n_prompt": 25,
166 + "n_gen": 128,
167 + "seconds": 2.3910184999695048,
168 + "text": "import csv\nfrom datetime import datetime\nfrom statistics import median\n\ndef median_inter_arrival_time(csv_file_path):\n "
169 + },
170 + {
171 + "name": "code_fit_2",
172 + "family": "code",
173 + "split": "fit",
174 + "prompt": "Implement an LRU cache class in Python with get and put methods in O(1).\n\n```pyt",
175 + "n_prompt": 21,
176 + "n_gen": 128,
177 + "seconds": 2.4063012920087203,
178 + "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self."
179 + },
180 + {
181 + "name": "code_fit_3",
182 + "family": "code",
183 + "split": "fit",
184 + "prompt": "Write a Python script that walks a directory tree and reports the total size per",
185 + "n_prompt": 21,
186 + "n_gen": 128,
187 + "seconds": 2.413348500034772,
188 + "text": "import os\nfrom collections import defaultdict\n\ndef walk_directory_tree(root_dir):\n \"\"\"\n Walks a directory tree and"
189 + },
190 + {
191 + "name": "code_fit_4",
192 + "family": "code",
193 + "split": "fit",
194 + "prompt": "Implement binary search over a sorted list of integers in Python and return the ",
195 + "n_prompt": 22,
196 + "n_gen": 128,
197 + "seconds": 2.4092803340172395,
198 + "text": "def binary_search_insert_index(arr, target):\n left, right = 0, len(arr)\n \n while left < right:\n mid = (l"
199 + },
200 + {
201 + "name": "code_fit_5",
202 + "family": "code",
203 + "split": "fit",
204 + "prompt": "Write a SQL query that returns the three most recent orders per customer from ta",
205 + "n_prompt": 27,
206 + "n_gen": 128,
207 + "seconds": 2.4406378749990836,
208 + "text": "WITH RankedOrders AS (\n SELECT \n customer_id,\n order_date,\n amount,\n ROW_NUMBER() OVER (P"
209 + },
210 + {
211 + "name": "code_fit_6",
212 + "family": "code",
213 + "split": "fit",
214 + "prompt": "Write a JavaScript function that debounces another function with a configurable ",
215 + "n_prompt": 17,
216 + "n_gen": 128,
217 + "seconds": 2.4138995829271153,
218 + "text": "function debounce(func, delay) {\n let timeoutId;\n\n return function (...args) {\n clearTimeout(timeoutId);\n\n timeo"
219 + },
220 + {
221 + "name": "code_fit_7",
222 + "family": "code",
223 + "split": "fit",
224 + "prompt": "Write a Rust function that counts word frequencies in a string and returns them ",
225 + "n_prompt": 21,
226 + "n_gen": 128,
227 + "seconds": 2.420720999944024,
228 + "text": "use std::collections::HashMap;\n\nfn count_word_frequencies(text: &str) -> Vec<(&str, usize)> {\n let mut word_count = H"
229 + },
230 + {
231 + "name": "code_fit_8",
232 + "family": "code",
233 + "split": "fit",
234 + "prompt": "class Matrix:\n \"\"\"Minimal dense matrix with multiplication and transpose.\"\"\"\n",
235 + "n_prompt": 22,
236 + "n_gen": 128,
237 + "seconds": 2.4007679170463234,
238 + "text": " self.rows = rows\n self.shape = (len(rows), len(rows[0]))\n\n def __mul__(self, other):\n assert se"
239 + },
240 + {
241 + "name": "code_fit_9",
242 + "family": "code",
243 + "split": "fit",
244 + "prompt": "Write a Python generator that yields sliding windows of size k over an iterable.",
245 + "n_prompt": 18,
246 + "n_gen": 128,
247 + "seconds": 2.4212146669160575,
248 + "text": "def sliding_window(iterable, k):\n \"\"\"\n Generator that yields sliding windows of size k over an iterable.\n\n Args"
249 + },
250 + {
251 + "name": "code_heldout_0",
252 + "family": "code",
253 + "split": "heldout",
254 + "prompt": "Write a C function that computes the dot product of two float arrays using NEON ",
255 + "n_prompt": 22,
256 + "n_gen": 128,
257 + "seconds": 2.419732499984093,
258 + "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n"
259 + },
260 + {
261 + "name": "code_heldout_1",
262 + "family": "code",
263 + "split": "heldout",
264 + "prompt": "Write a bash script that lists the ten largest files under a directory.\n\n```bash",
265 + "n_prompt": 17,
266 + "n_gen": 128,
267 + "seconds": 2.396812416962348,
268 + "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exi"
269 + },
270 + {
271 + "name": "code_heldout_2",
272 + "family": "code",
273 + "split": "heldout",
274 + "prompt": "Implement Dijkstra's shortest path in Python using a binary heap.\n\n```python\n",
275 + "n_prompt": 16,
276 + "n_gen": 128,
277 + "seconds": 2.4002217499073595,
278 + "text": "import heapq\nfrom collections import defaultdict\nimport sys\n\ndef dijkstra(graph, start, end):\n # Initialize distances"
279 + },
280 + {
281 + "name": "code_heldout_3",
282 + "family": "code",
283 + "split": "heldout",
284 + "prompt": "Write a Go function that reads lines from stdin and prints the longest one.\n\n```",
285 + "n_prompt": 18,
286 + "n_gen": 128,
287 + "seconds": 2.3987378750462085,
288 + "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)"
289 + },
290 + {
291 + "name": "reasoning_fit_0",
292 + "family": "reasoning",
293 + "split": "fit",
294 + "prompt": "Q: If a train travels 60 km in 45 minutes, what is its average speed in km/h? Th",
295 + "n_prompt": 32,
296 + "n_gen": 128,
297 + "seconds": 2.462184250005521,
298 + "text": " To find the average speed, we use the formula: speed = distance / time. The distance is 60 km, and the time is 45 minut"
299 + },
300 + {
301 + "name": "reasoning_fit_1",
302 + "family": "reasoning",
303 + "split": "fit",
304 + "prompt": "Q: A rectangle has a perimeter of 36 cm and a length twice its width. What is it",
305 + "n_prompt": 31,
306 + "n_gen": 128,
307 + "seconds": 2.396634125034325,
308 + "text": " Let the width be w and the length be 2w. The perimeter is 2(w + 2w) = 36 cm. So, 2(3w) = 36 cm, which gives 6w = 36 cm,"
309 + },
310 + {
311 + "name": "reasoning_fit_2",
312 + "family": "reasoning",
313 + "split": "fit",
314 + "prompt": "Q: Three friends split a bill of 84 dollars; one pays twice as much as each of t",
315 + "n_prompt": 37,
316 + "n_gen": 128,
317 + "seconds": 2.4701771249528974,
318 + "text": " Let x be the amount each of the two friends pays. Then the third friend pays 2x. The total bill is x + x + 2x = 84. So "
319 + },
320 + {
321 + "name": "reasoning_fit_3",
322 + "family": "reasoning",
323 + "split": "fit",
324 + "prompt": "Q: A store sells pens at 3 for 2 dollars. How much do 18 pens cost? Think step b",
325 + "n_prompt": 30,
326 + "n_gen": 128,
327 + "seconds": 2.437447457923554,
328 + "text": " 3 pens cost $2, so 18 pens is 18 / 3 = 6 sets of pens. Each set costs $2, so 6 sets cost 6 * $2 = $12. The answer is $1"
329 + },
330 + {
331 + "name": "reasoning_fit_4",
332 + "family": "reasoning",
333 + "split": "fit",
334 + "prompt": "Q: If today is Wednesday, what day of the week will it be in 100 days? Think ste",
335 + "n_prompt": 29,
336 + "n_gen": 128,
337 + "seconds": 2.4229157500667498,
338 + "text": " There are 7 days in a week, so the days of the week repeat every 7 days. To find out what day it will be in 100 days, w"
339 + },
340 + {
341 + "name": "reasoning_fit_5",
342 + "family": "reasoning",
343 + "split": "fit",
344 + "prompt": "Q: A car uses 7 litres per 100 km. How many litres for a 350 km trip, and the co",
345 + "n_prompt": 46,
346 + "n_gen": 128,
347 + "seconds": 2.453943208907731,
348 + "text": " Step 1: Determine the fuel consumption rate\nThe car uses 7 litres per 100 km.\n\nStep 2: Calculate fuel needed for 350 km"
349 + },
350 + {
351 + "name": "reasoning_fit_6",
352 + "family": "reasoning",
353 + "split": "fit",
354 + "prompt": "Q: The sum of two consecutive even numbers is 54. What are they? Think step by s",
355 + "n_prompt": 25,
356 + "n_gen": 128,
357 + "seconds": 2.403570582973771,
358 + "text": " Let the two consecutive even numbers be x and x + 2. Then, x + (x + 2) = 54. This simplifies to 2x + 2 = 54. Subtract 2"
359 + },
360 + {
361 + "name": "reasoning_fit_7",
362 + "family": "reasoning",
363 + "split": "fit",
364 + "prompt": "Q: A recipe for 4 people needs 300 g of flour. How much for 10 people? Think ste",
365 + "n_prompt": 32,
366 + "n_gen": 128,
367 + "seconds": 2.4298152920091525,
368 + "text": " First find how much flour is needed per person: 300 g / 4 people = 75 g per person. Then multiply that amount by 10 peo"
369 + },
370 + {
371 + "name": "reasoning_fit_8",
372 + "family": "reasoning",
373 + "split": "fit",
374 + "prompt": "Q: A clock shows 3:15. What is the angle between the hour and minute hands? Thin",
375 + "n_prompt": 29,
376 + "n_gen": 128,
377 + "seconds": 2.4581131249433383,
378 + "text": " Let's think step by step.\n\n1. First, determine the position of the minute hand:\n - At 15 minutes past the hour, the m"
379 + },
380 + {
381 + "name": "reasoning_fit_9",
382 + "family": "reasoning",
383 + "split": "fit",
384 + "prompt": "Q: If 5 machines make 5 widgets in 5 minutes, how long do 100 machines take to m",
385 + "n_prompt": 39,
386 + "n_gen": 128,
387 + "seconds": 2.471293666982092,
388 + "text": " Let's break this down step by step.\n\nWe are told:\n- 5 machines make 5 widgets in 5 minutes.\n\nStep 1: Determine how long"
389 + },
390 + {
391 + "name": "reasoning_heldout_0",
392 + "family": "reasoning",
393 + "split": "heldout",
394 + "prompt": "Q: A tank fills in 6 hours with pipe A and 9 hours with pipe B. How long with bo",
395 + "n_prompt": 33,
396 + "n_gen": 128,
397 + "seconds": 2.451238750014454,
398 + "text": " Pipe A fills 1/6 of the tank per hour. Pipe B fills 1/9 of the tank per hour. Together they fill 1/6 + 1/9 = 5/18 of th"
399 + },
400 + {
401 + "name": "reasoning_heldout_1",
402 + "family": "reasoning",
403 + "split": "heldout",
404 + "prompt": "Q: What is the smallest positive integer divisible by 2, 3, 4, 5 and 6? Think st",
405 + "n_prompt": 32,
406 + "n_gen": 128,
407 + "seconds": 2.408647916978225,
408 + "text": " We want to find the smallest positive integer divisible by 2, 3, 4, 5, and 6.\n\nStep 1: Find the least common multiple ("
409 + },
410 + {
411 + "name": "reasoning_heldout_2",
412 + "family": "reasoning",
413 + "split": "heldout",
414 + "prompt": "Q: A shirt costs 40 dollars after a 20 percent discount. What was the original p",
415 + "n_prompt": 30,
416 + "n_gen": 128,
417 + "seconds": 2.412208624999039,
418 + "text": " Let x be the original price. Then x - 0.2x = 0.8x = 40. So x = 40 / 0.8 = 50 dollars. The answer is 50 dollars.\n\nQ: A s"
419 + },
420 + {
421 + "name": "reasoning_heldout_3",
422 + "family": "reasoning",
423 + "split": "heldout",
424 + "prompt": "Q: Two dice are rolled. What is the probability that the sum is 7? Think step by",
425 + "n_prompt": 25,
426 + "n_gen": 128,
427 + "seconds": 2.4018902089446783,
428 + "text": " To find the probability that the sum of two dice is 7, let's go step by step:\n\nStep 1: Determine the total number of po"
429 + },
430 + {
431 + "name": "multilingual_fit_0",
432 + "family": "multilingual",
433 + "split": "fit",
434 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent seulement une fraction de leurs param\u00e8t",
435 + "n_prompt": 22,
436 + "n_gen": 128,
437 + "seconds": 2.41435641690623,
438 + "text": " chaque expert ne traite qu'une partie du probl\u00e8me.\n\nR\u00e9ponse:\nNone\n\nQuel est le r\u00f4le des algorithmes de recherche dans l"
439 + },
440 + {
441 + "name": "multilingual_fit_1",
442 + "family": "multilingual",
443 + "split": "fit",
444 + "prompt": "Explique en fran\u00e7ais pourquoi la m\u00e9moire unifi\u00e9e des puces Apple est utile pour ",
445 + "n_prompt": 32,
446 + "n_gen": 128,
447 + "seconds": 2.4370479590725154,
448 + "text": " La m\u00e9moire unifi\u00e9e des puces Apple, comme celle utilis\u00e9e dans les puce M1, M2, etc., est particuli\u00e8rement utile pour l'"
449 + },
450 + {
451 + "name": "multilingual_fit_2",
452 + "family": "multilingual",
453 + "split": "fit",
454 + "prompt": "R\u00e9dige un court paragraphe sur l'histoire de la ville de Qu\u00e9bec.",
455 + "n_prompt": 19,
456 + "n_gen": 128,
457 + "seconds": 2.3830997500335798,
458 + "text": " La ville de Qu\u00e9bec, fond\u00e9e en 1608 par Samuel de Champlain, est la plus ancienne cit\u00e9 continue en Am\u00e9rique du Nord fond"
459 + },
460 + {
461 + "name": "multilingual_fit_3",
462 + "family": "multilingual",
463 + "split": "fit",
464 + "prompt": "Quels sont les avantages du transport en commun dans une ville de taille moyenne",
465 + "n_prompt": 24,
466 + "n_gen": 128,
467 + "seconds": 2.411749624996446,
468 + "text": " Le transport en commun dans une ville de taille moyenne offre de nombreux avantages, tant pour les citoyens que pour l\u2019"
469 + },
470 + {
471 + "name": "multilingual_fit_4",
472 + "family": "multilingual",
473 + "split": "fit",
474 + "prompt": "Erkl\u00e4re in drei S\u00e4tzen, warum Fahrr\u00e4der in St\u00e4dten n\u00fctzlich sind.",
475 + "n_prompt": 25,
476 + "n_gen": 128,
477 + "seconds": 2.4059482091106474,
478 + "text": " Fahrr\u00e4der sind umweltfreundlich, da sie keine Emissionen verursachen und somit die Luftqualit\u00e4t in St\u00e4dten verbessern. "
479 + },
480 + {
481 + "name": "multilingual_fit_5",
482 + "family": "multilingual",
483 + "split": "fit",
484 + "prompt": "Describe en espa\u00f1ol c\u00f3mo se prepara un buen caf\u00e9 con una cafetera italiana.",
485 + "n_prompt": 17,
486 + "n_gen": 128,
487 + "seconds": 2.4173996669705957,
488 + "text": " \u00a1Claro! Aqu\u00ed tienes una descripci\u00f3n en espa\u00f1ol de c\u00f3mo preparar un buen caf\u00e9 con una cafetera italiana (tambi\u00e9n conocid"
489 + },
490 + {
491 + "name": "multilingual_fit_6",
492 + "family": "multilingual",
493 + "split": "fit",
494 + "prompt": "Explique en fran\u00e7ais la diff\u00e9rence entre un pr\u00eat hypoth\u00e9caire \u00e0 taux fixe et \u00e0 t",
495 + "n_prompt": 24,
496 + "n_gen": 128,
497 + "seconds": 2.3836591669823974,
498 + "text": " Un pr\u00eat hypoth\u00e9caire \u00e0 taux fixe est un pr\u00eat o\u00f9 le taux d'int\u00e9r\u00eat reste constant pendant toute la dur\u00e9e du pr\u00eat. Cela s"
499 + },
500 + {
501 + "name": "multilingual_fit_7",
502 + "family": "multilingual",
503 + "split": "fit",
504 + "prompt": "Scrivi un breve paragrafo sulla cucina tradizionale della Toscana.",
505 + "n_prompt": 18,
506 + "n_gen": 128,
507 + "seconds": 2.4004312499891967,
508 + "text": " La cucina tradizionale della Toscana \u00e8 un esempio di semplicit\u00e0 e qualit\u00e0, fondata su ingredienti freschi e di origine "
509 + },
510 + {
511 + "name": "multilingual_heldout_0",
512 + "family": "multilingual",
513 + "split": "heldout",
514 + "prompt": "D\u00e9cris en quelques phrases le fonctionnement d'un cache LRU.",
515 + "n_prompt": 15,
516 + "n_gen": 128,
517 + "seconds": 2.4136952919652686,
518 + "text": " Un cache LRU (Least Recently Used) est une strat\u00e9gie de gestion de m\u00e9moire cache qui supprime l'\u00e9l\u00e9ment le moins r\u00e9cemm"
519 + },
520 + {
521 + "name": "multilingual_heldout_1",
522 + "family": "multilingual",
523 + "split": "heldout",
524 + "prompt": "\u00bfCu\u00e1les son las principales causas del cambio clim\u00e1tico? Responde brevemente en ",
525 + "n_prompt": 22,
526 + "n_gen": 128,
527 + "seconds": 2.3792647500522435,
528 + "text": " Las principales causas del cambio clim\u00e1tico son la emisi\u00f3n de gases de efecto invernadero, como di\u00f3xido de carbono y me"
529 + },
530 + {
531 + "name": "multilingual_heldout_2",
532 + "family": "multilingual",
533 + "split": "heldout",
534 + "prompt": "Beschreibe kurz, wie ein Mietvertrag in Deutschland funktioniert.",
535 + "n_prompt": 17,
536 + "n_gen": 128,
537 + "seconds": 2.409194999956526,
538 + "text": " Ein Mietvertrag in Deutschland ist ein rechtlich verbindliches Dokument zwischen Vermieter und Mieter, das die Bedingun"
539 + },
540 + {
541 + "name": "longctx_fit_0",
542 + "family": "longctx",
543 + "split": "fit",
544 + "prompt": "You are an assistant helping an economics professor. Context: the course covers ",
545 + "n_prompt": 136,
546 + "n_gen": 128,
547 + "seconds": 2.6195002499734983,
548 + "text": " The lab should include a step-by-step guide, a dataset description, and a discussion prompt.\n\n# **Two-Hour Lab: Estimat"
549 + },
550 + {
551 + "name": "longctx_fit_1",
552 + "family": "longctx",
553 + "split": "fit",
554 + "prompt": "Context: a small laboratory operates a cluster of twenty-nine Apple Silicon comp",
555 + "n_prompt": 100,
556 + "n_gen": 128,
557 + "seconds": 2.531550707994029,
558 + "text": " \n\n# **Comprehensive Maintenance Checklist for Apple Silicon Lab Cluster**\n\n## **1. Daily Tasks**\n\n| Task | Justificatio"
559 + },
560 + {
561 + "name": "longctx_fit_2",
562 + "family": "longctx",
563 + "split": "fit",
564 + "prompt": "Context: a municipality in Quebec is revising its property assessment roll. Curr",
565 + "n_prompt": 92,
566 + "n_gen": 128,
567 + "seconds": 2.51342058298178,
568 + "text": " Use the following information: 1. The municipality has a database of 10,000 properties, including sale prices, property"
569 + },
570 + {
571 + "name": "longctx_heldout_0",
572 + "family": "longctx",
573 + "split": "heldout",
574 + "prompt": "Context: a graduate student is designing a benchmark to compare inference runtim",
575 + "n_prompt": 103,
576 + "n_gen": 128,
577 + "seconds": 2.5320626670727506,
578 + "text": " The goal is to produce a benchmark that is both rigorous and practical for future researchers.\n\nWe are designing a benc"
579 + },
580 + {
581 + "name": "drift_fit_0",
582 + "family": "drift",
583 + "split": "fit",
584 + "prompt": "First, write a Python function that reverses a linked list.\n\n```python\n",
585 + "n_prompt": 16,
586 + "n_gen": 128,
587 + "seconds": 2.3907273749355227,
588 + "text": "class ListNode:\n def __init__(self, val=0, next=None):\n self.val = val\n self.next = next\n\ndef reverse_l"
589 + },
590 + {
591 + "name": "drift_fit_1",
592 + "family": "drift",
593 + "split": "fit",
594 + "prompt": "Explain the rules of ice hockey in two paragraphs, then write a SQL query counti",
595 + "n_prompt": 28,
596 + "n_gen": 128,
597 + "seconds": 2.4191840830026194,
598 + "text": " Ice hockey is a fast-paced team sport played on ice, where two teams of six players each (including a goalie) compete t"
599 + },
600 + {
601 + "name": "drift_fit_2",
602 + "family": "drift",
603 + "split": "fit",
604 + "prompt": "R\u00e9sume en fran\u00e7ais l'intrigue d'un roman policier imaginaire, puis donne la solu",
605 + "n_prompt": 39,
606 + "n_gen": 128,
607 + "seconds": 2.4495359159773216,
608 + "text": " Enfin, \u00e9cris une courte description d'un paysage de montagne au cr\u00e9puscule.\n\n**R\u00e9sum\u00e9 du roman policier imaginaire :** "
609 + },
610 + {
611 + "name": "drift_heldout_0",
612 + "family": "drift",
613 + "split": "heldout",
614 + "prompt": "Describe a sunset over the Ottawa River in one paragraph, then implement quickso",
615 + "n_prompt": 21,
616 + "n_gen": 128,
617 + "seconds": 2.402475666021928,
618 + "text": "The Ottawa River glows in the final light of day as the sun sinks behind the distant trees of Gatineau Park, painting th"
619 + },
620 + {
621 + "name": "drift_heldout_1",
622 + "family": "drift",
623 + "split": "heldout",
624 + "prompt": "Q: What is 17 times 23? Think step by step.\nA: Then write a haiku about autumn i",
625 + "n_prompt": 29,
626 + "n_gen": 128,
627 + "seconds": 2.435008665896021,
628 + "text": " Think step by step.\n\nA: 17 \u00d7 23 \nBreak it down: 17 \u00d7 (20 + 3) \n= (17 \u00d7 20) + (17 \u00d7 3) \n= 340 + 51 \n= 391 \n\nAnswer:"
629 + }
630 + ]
631 +}
\ No newline at end of file
added benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b_v1_fp16wire.jsonl +144 −0
@@ -0,0 +1,144 @@
1 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:31:07", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1305.0666250055656, "prompt_tps": 16.881093863542993, "gen_tps": 13.153067912446549, "decode_tps_wall": 12.947744144622982, "itl_ms_p50": 65.32775005325675, "itl_ms_p95": 172.31916694436222, "itl_ms_p99": 207.94133306480944, "itl_ms_mean": 77.23353109470317, "total_s": 6.1707790839718655, "peak_mem_gb": 5.387600848, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793514661.268065, "t_end_epoch": 1793514667.4401128, "t_first_epoch": 1793514662.5731316, "cache_hit_rate": 0.8028462339465463, "cache_misses": 5680, "stall_s": 3.338298593647778, "remote": null}
2 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:31:14", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 840.5029170680791, "prompt_tps": 16.67297939384595, "gen_tps": 9.840121851526153, "decode_tps_wall": 9.686387881754953, "itl_ms_p50": 89.53474997542799, "itl_ms_p95": 174.8542080167681, "itl_ms_p99": 225.49499990418553, "itl_ms_mean": 103.23765806277238, "total_s": 7.344475375022739, "peak_mem_gb": 5.262111634, "text": "The Rideau Canal, located in Ontario, Canada, was constructed between 1826 and 1832 as a strategic military waterway to provide a secure alternative route between Montreal and Lake Ontario, bypassing ", "t_start_epoch": 1793514667.440329, "t_end_epoch": 1793514674.786064, "t_first_epoch": 1793514668.280832, "cache_hit_rate": 0.7851206821795057, "cache_misses": 5947, "stall_s": 3.6807542687747627, "remote": null}
3 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:31:23", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 841.6604170342907, "prompt_tps": 20.213477100358563, "gen_tps": 8.586669094086856, "decode_tps_wall": 8.452523322921468, "itl_ms_p50": 120.05683302413672, "itl_ms_p95": 186.77829205989838, "itl_ms_p99": 211.56074991449714, "itl_ms_mean": 118.30786639632332, "total_s": 8.29505600000266, "peak_mem_gb": 5.275638676, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793514674.786211, "t_end_epoch": 1793514683.083704, "t_first_epoch": 1793514675.6278715, "cache_hit_rate": 0.7559745747665211, "cache_misses": 6872, "stall_s": 4.421856662724167, "remote": null}
4 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:31:30", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 831.0396670131013, "prompt_tps": 19.298992235277943, "gen_tps": 9.380738830897483, "decode_tps_wall": 9.234162531883733, "itl_ms_p50": 108.2873340928927, "itl_ms_p95": 173.58050006441772, "itl_ms_p99": 238.67691599298269, "itl_ms_mean": 108.29352380870472, "total_s": 7.653531666961499, "peak_mem_gb": 5.249805988, "text": "Include a sentence about the beauty of music and a sentence about perseverance.\n\n**Dear Beginner,**\n\nStarting to learn the piano as an adult is one of the most beautiful gifts you can give yourself. M", "t_start_epoch": 1793514683.084123, "t_end_epoch": 1793514690.740486, "t_first_epoch": 1793514683.9151626, "cache_hit_rate": 0.815905208446916, "cache_misses": 5065, "stall_s": 3.6146102038910612, "remote": null}
5 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:31:38", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 836.4636249607429, "prompt_tps": 19.218242910345122, "gen_tps": 8.79065620556938, "decode_tps_wall": 8.653301805243345, "itl_ms_p50": 114.6399580175057, "itl_ms_p95": 202.99820799846202, "itl_ms_p99": 226.7304580891505, "itl_ms_mean": 115.56282474674167, "total_s": 8.116921584005468, "peak_mem_gb": 5.252212336, "text": "The Mac Studio and MacBook Pro are both high-performance Apple devices designed for demanding workloads, including scientific computing. However, they differ significantly in form factor, performance,", "t_start_epoch": 1793514690.741142, "t_end_epoch": 1793514698.8608398, "t_first_epoch": 1793514691.5776057, "cache_hit_rate": 0.7660601693388579, "cache_misses": 6493, "stall_s": 4.256898548803292, "remote": null}
6 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:31:46", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 859.00445911102, "prompt_tps": 17.533453227223788, "gen_tps": 9.204380923835247, "decode_tps_wall": 9.060565674853445, "itl_ms_p50": 108.94774994812906, "itl_ms_p95": 175.8351669413969, "itl_ms_p99": 257.55908305291086, "itl_ms_mean": 110.36838492053369, "total_s": 7.812212709104642, "peak_mem_gb": 5.256847388, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety precautions should hikers take, and what are the benefits of hiking ", "t_start_epoch": 1793514698.861375, "t_end_epoch": 1793514706.676335, "t_first_epoch": 1793514699.7203796, "cache_hit_rate": 0.7615640358485112, "cache_misses": 6598, "stall_s": 4.144967816071585, "remote": null}
7 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:31:54", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 922.8450420778245, "prompt_tps": 19.568739404040024, "gen_tps": 9.43444419535548, "decode_tps_wall": 9.287041558864363, "itl_ms_p50": 108.48862503189594, "itl_ms_p95": 183.08604205958545, "itl_ms_p99": 210.70466691162437, "itl_ms_mean": 107.67691666519062, "total_s": 7.706490791984834, "peak_mem_gb": 5.291424328, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793514706.6769001, "t_end_epoch": 1793514714.385064, "t_first_epoch": 1793514707.5997453, "cache_hit_rate": 0.821324330977248, "cache_misses": 5081, "stall_s": 3.660260966862552, "remote": null}
8 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:32:02", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 827.7329590637237, "prompt_tps": 19.369028662536177, "gen_tps": 8.374168078612534, "decode_tps_wall": 8.243328085692195, "itl_ms_p50": 127.29495798703283, "itl_ms_p95": 184.84820809680969, "itl_ms_p99": 207.09908404387534, "itl_ms_mean": 121.3102268409871, "total_s": 8.47027725004591, "peak_mem_gb": 5.294334792, "text": "import heapq\nfrom collections import defaultdict\nimport sys\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = {node: float('inf') for node in graph}\n distance", "t_start_epoch": 1793514714.38537, "t_end_epoch": 1793514722.858221, "t_first_epoch": 1793514715.213103, "cache_hit_rate": 0.7550634716873484, "cache_misses": 6869, "stall_s": 4.591691055917181, "remote": null}
9 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:32:10", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 847.1207499969751, "prompt_tps": 16.567731981198495, "gen_tps": 8.968149162010402, "decode_tps_wall": 8.828023584115142, "itl_ms_p50": 110.42537505272776, "itl_ms_p95": 180.03325001336634, "itl_ms_p99": 228.14591601490974, "itl_ms_mean": 113.27563757298603, "total_s": 7.983485917095095, "peak_mem_gb": 5.259932684, "text": "The Rideau Canal, located in Ontario, Canada, was constructed between 1826 and 1832 as a strategic military waterway to provide a secure alternative route between Montreal and Lake Ontario, bypassing ", "t_start_epoch": 1793514722.8587148, "t_end_epoch": 1793514730.844877, "t_first_epoch": 1793514723.7058356, "cache_hit_rate": 0.7830611360023124, "cache_misses": 6004, "stall_s": 4.014461511280388, "remote": null}
10 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:32:18", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 843.6830829596147, "prompt_tps": 19.032084675420844, "gen_tps": 8.811137635959831, "decode_tps_wall": 8.673467491219354, "itl_ms_p50": 116.961708990857, "itl_ms_p95": 204.63204104453325, "itl_ms_p99": 243.36537497583777, "itl_ms_mean": 115.29414285720873, "total_s": 8.107214082963765, "peak_mem_gb": 5.243282104, "text": "The Mac Studio and MacBook Pro are both high-performance Apple devices designed for demanding workloads, including scientific computing. However, they differ significantly in form factor, performance,", "t_start_epoch": 1793514730.845443, "t_end_epoch": 1793514738.955709, "t_first_epoch": 1793514731.689126, "cache_hit_rate": 0.7659881102504054, "cache_misses": 6495, "stall_s": 4.3121169480727986, "remote": null}
11 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:32:26", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 953.1139169121161, "prompt_tps": 23.149796352454395, "gen_tps": 10.147263415986863, "decode_tps_wall": 9.98872885613137, "itl_ms_p50": 97.24941698368639, "itl_ms_p95": 177.29304207023233, "itl_ms_p99": 199.12399991881102, "itl_ms_mean": 100.1128386207191, "total_s": 7.2602227500174195, "peak_mem_gb": 5.3955144, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793514738.9562528, "t_end_epoch": 1793514746.217748, "t_first_epoch": 1793514739.9093668, "cache_hit_rate": 0.7965636931620965, "cache_misses": 5861, "stall_s": 3.531949420575984, "remote": null}
12 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:32:34", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 793.4672910487279, "prompt_tps": 21.441422778453237, "gen_tps": 8.411677576769486, "decode_tps_wall": 8.280254999561981, "itl_ms_p50": 119.42516593262553, "itl_ms_p95": 185.59299991466105, "itl_ms_p99": 210.4800830129534, "itl_ms_mean": 120.76922752414016, "total_s": 8.401928625069559, "peak_mem_gb": 5.285747048, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793514746.217885, "t_end_epoch": 1793514754.621513, "t_first_epoch": 1793514747.0113523, "cache_hit_rate": 0.7616917012890168, "cache_misses": 6711, "stall_s": 4.425078197848052, "remote": null}
13 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:32:42", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 868.2223340729252, "prompt_tps": 17.305341941852163, "gen_tps": 8.570224819090752, "decode_tps_wall": 8.436324941231847, "itl_ms_p50": 111.6302500013262, "itl_ms_p95": 192.82016705255955, "itl_ms_p99": 385.16237505245954, "itl_ms_mean": 118.53502644410742, "total_s": 8.335929000051692, "peak_mem_gb": 5.26795574, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety precautions should hikers take, and what are the benefits of hiking ", "t_start_epoch": 1793514754.621831, "t_end_epoch": 1793514762.960455, "t_first_epoch": 1793514755.4900532, "cache_hit_rate": 0.7535414859786065, "cache_misses": 6820, "stall_s": 4.831829619943164, "remote": null}
14 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:32:49", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 695.6080830423161, "prompt_tps": 23.089875042299656, "gen_tps": 11.106519789578048, "decode_tps_wall": 10.932985872599767, "itl_ms_p50": 83.03079195320606, "itl_ms_p95": 149.87716602627188, "itl_ms_p99": 232.1509999455884, "itl_ms_mean": 91.4663214288238, "total_s": 6.457986333058216, "peak_mem_gb": 5.207338656, "text": "Include a sentence about the beauty of music and a sentence about perseverance.\n\n**Dear Beginner,**\n\nStarting to learn the piano as an adult is one of the most beautiful gifts you can give yourself. M", "t_start_epoch": 1793514762.9609869, "t_end_epoch": 1793514769.42111, "t_first_epoch": 1793514763.656595, "cache_hit_rate": 0.8319339948388035, "cache_misses": 4624, "stall_s": 3.0101614736486226, "remote": null}
15 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:32:57", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 879.6437919372693, "prompt_tps": 20.529731726237603, "gen_tps": 9.22076874243682, "decode_tps_wall": 9.076704255612038, "itl_ms_p50": 108.5839580046013, "itl_ms_p95": 176.98912508785725, "itl_ms_p99": 205.63583297189325, "itl_ms_mean": 110.17214749304075, "total_s": 7.820489083998837, "peak_mem_gb": 5.291424328, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793514769.421629, "t_end_epoch": 1793514777.243925, "t_first_epoch": 1793514770.3012726, "cache_hit_rate": 0.8207265182684531, "cache_misses": 5098, "stall_s": 3.6823155300226063, "remote": null}
16 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:33:06", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 854.2657920625061, "prompt_tps": 18.76654994938428, "gen_tps": 8.125298923462005, "decode_tps_wall": 7.998348447136505, "itl_ms_p50": 128.30125004984438, "itl_ms_p95": 187.7852079924196, "itl_ms_p99": 254.09824994858354, "itl_ms_mean": 125.02581084198836, "total_s": 8.730891875107773, "peak_mem_gb": 5.294334792, "text": "import heapq\nfrom collections import defaultdict\nimport sys\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = {node: float('inf') for node in graph}\n distance", "t_start_epoch": 1793514777.244253, "t_end_epoch": 1793514786.044668, "t_first_epoch": 1793514778.0985186, "cache_hit_rate": 0.7550634716873484, "cache_misses": 6869, "stall_s": 4.727746542659588, "remote": null}
17 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:33:14", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 833.6901670554653, "prompt_tps": 20.423757921627345, "gen_tps": 7.952405186994998, "decode_tps_wall": 7.828158744367814, "itl_ms_p50": 129.10245801322162, "itl_ms_p95": 191.00791693199426, "itl_ms_p99": 216.3085830397904, "itl_ms_mean": 127.7439603175495, "total_s": 8.881559667061083, "peak_mem_gb": 5.288699444, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793514786.0449789, "t_end_epoch": 1793514794.927948, "t_first_epoch": 1793514786.878669, "cache_hit_rate": 0.7594900749263165, "cache_misses": 6773, "stall_s": 4.71670598781202, "remote": null}
18 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:33:23", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 945.6842080689967, "prompt_tps": 23.2782842888627, "gen_tps": 8.928762739262817, "decode_tps_wall": 8.789257206347209, "itl_ms_p50": 107.74579201824963, "itl_ms_p95": 209.27395799662918, "itl_ms_p99": 219.77658302057534, "itl_ms_mean": 113.77525728543304, "total_s": 8.113525417051278, "peak_mem_gb": 5.401216032, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793514794.928097, "t_end_epoch": 1793514803.0439832, "t_first_epoch": 1793514795.8737812, "cache_hit_rate": 0.7960777507809789, "cache_misses": 5875, "stall_s": 4.0277121644467115, "remote": null}
19 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:33:31", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 833.7812500540167, "prompt_tps": 18.009439662440712, "gen_tps": 8.471042510517156, "decode_tps_wall": 8.338689462230327, "itl_ms_p50": 123.11316700652242, "itl_ms_p95": 168.54220803361386, "itl_ms_p99": 262.6902920892462, "itl_ms_mean": 119.92292128510714, "total_s": 8.388925291015767, "peak_mem_gb": 5.254160412, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety precautions should hikers take, and what are the benefits of hiking ", "t_start_epoch": 1793514803.044199, "t_end_epoch": 1793514811.43616, "t_first_epoch": 1793514803.8779802, "cache_hit_rate": 0.7580948250939578, "cache_misses": 6694, "stall_s": 4.471358717069961, "remote": null}
20 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:33:40", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 858.5433750413358, "prompt_tps": 18.699685246933957, "gen_tps": 8.092164655388402, "decode_tps_wall": 7.965733773220115, "itl_ms_p50": 128.63587494939566, "itl_ms_p95": 201.2748330598697, "itl_ms_p99": 220.8289159461856, "itl_ms_mean": 125.53771296774761, "total_s": 8.767419292009436, "peak_mem_gb": 5.276149364, "text": "import heapq\nfrom collections import defaultdict\nimport sys\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = {node: float('inf') for node in graph}\n distance", "t_start_epoch": 1793514811.436766, "t_end_epoch": 1793514820.2061079, "t_first_epoch": 1793514812.2953093, "cache_hit_rate": 0.7551704464413065, "cache_misses": 6866, "stall_s": 4.700590412248857, "remote": null}
21 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:33:48", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 851.583666051738, "prompt_tps": 18.822983166515776, "gen_tps": 8.929302405518488, "decode_tps_wall": 8.789790741772896, "itl_ms_p50": 107.82079200726002, "itl_ms_p95": 182.4391249101609, "itl_ms_p99": 224.8037919634953, "itl_ms_mean": 113.76835119038347, "total_s": 8.018989791045897, "peak_mem_gb": 5.273101424, "text": "The Mac Studio and MacBook Pro are both high-performance Apple devices designed for demanding workloads, including scientific computing. However, they differ significantly in form factor, performance,", "t_start_epoch": 1793514820.206418, "t_end_epoch": 1793514828.227509, "t_first_epoch": 1793514821.0580018, "cache_hit_rate": 0.7578094037110431, "cache_misses": 6722, "stall_s": 4.235948825720698, "remote": null}
22 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:33:56", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 758.8960829889402, "prompt_tps": 18.50271817585031, "gen_tps": 8.801383275004348, "decode_tps_wall": 8.663866228202146, "itl_ms_p50": 113.97029203362763, "itl_ms_p95": 177.64358292333782, "itl_ms_p99": 201.80929196067154, "itl_ms_mean": 115.42191138003199, "total_s": 8.030476499930955, "peak_mem_gb": 5.24353221, "text": "The Rideau Canal, located in Ontario, Canada, was constructed between 1826 and 1832 as a strategic military waterway to provide a secure alternative route between Montreal and Lake Ontario, bypassing ", "t_start_epoch": 1793514828.2279701, "t_end_epoch": 1793514836.2606761, "t_first_epoch": 1793514828.9868662, "cache_hit_rate": 0.7924194247723659, "cache_misses": 5745, "stall_s": 4.029636505409144, "remote": null}
23 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:34:04", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 889.9925840087235, "prompt_tps": 20.28689901661177, "gen_tps": 8.898557366866203, "decode_tps_wall": 8.759521417403903, "itl_ms_p50": 117.08283401094377, "itl_ms_p95": 183.79762500990182, "itl_ms_p99": 210.91337502002716, "itl_ms_mean": 114.161488093761, "total_s": 8.082166333915666, "peak_mem_gb": 5.286121578, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793514836.261167, "t_end_epoch": 1793514844.34606, "t_first_epoch": 1793514837.1511595, "cache_hit_rate": 0.8197770510250729, "cache_misses": 5125, "stall_s": 3.769628528621979, "remote": null}
24 +{"backend": "titan-distributed", "capacity": 32, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5044159479439259, 0.5606659688055515, 0.5571249639615417, 0.5678749876096845, 0.5484579596668482, 0.5486250156536698, 0.5195410922169685, 0.5602500168606639, 0.5976660177111626, 0.5560829304158688]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:34:12", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 815.7977079972625, "prompt_tps": 19.672915233149567, "gen_tps": 9.130565933759, "decode_tps_wall": 8.987903299150785, "itl_ms_p50": 106.87404102645814, "itl_ms_p95": 183.7390410946682, "itl_ms_p99": 230.0831669708714, "itl_ms_mean": 111.26065409431855, "total_s": 7.825218915939331, "peak_mem_gb": 5.236534916, "text": "Include a sentence about the beauty of music and a sentence about perseverance.\n\n**Dear Beginner,**\n\nStarting to learn the piano as an adult is one of the most beautiful gifts you can give yourself. M", "t_start_epoch": 1793514844.3465889, "t_end_epoch": 1793514852.1750379, "t_first_epoch": 1793514845.1623867, "cache_hit_rate": 0.816195980082143, "cache_misses": 5057, "stall_s": 3.7392752155428752, "remote": null}
25 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:34:41", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1579.2659589787945, "prompt_tps": 15.036986287182895, "gen_tps": 3.9205977971042394, "decode_tps_wall": 3.859363881835501, "itl_ms_p50": 259.7795419860631, "itl_ms_p95": 286.6639170097187, "itl_ms_p99": 294.3278750171885, "itl_ms_mean": 259.11005819031584, "total_s": 17.903199624968693, "peak_mem_gb": 5.29792424, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793514863.4970722, "t_end_epoch": 1793514881.402548, "t_first_epoch": 1793514865.0763383, "cache_hit_rate": 0.8841207917930934, "cache_misses": 1604, "stall_s": 1.4832589607685804, "remote": {"remote_pairs": 17174, "local_pairs": 15850, "layers_remote": 3257, "remote_wait_s": 8.291328542749397, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 3306, "exec_wall_s": 10.601246706908569, "bytes_sent": 18324112, "bytes_recv": 74509253}}}}
26 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:34:59", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 1167.4354580463842, "prompt_tps": 12.01010778927316, "gen_tps": 3.775092865390118, "decode_tps_wall": 3.7161210588632234, "itl_ms_p50": 274.8935829149559, "itl_ms_p95": 290.2214580681175, "itl_ms_p99": 306.4427920617163, "itl_ms_mean": 269.09779960341336, "total_s": 18.120596833061427, "peak_mem_gb": 5.167307556, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793514881.402966, "t_end_epoch": 1793514899.527565, "t_first_epoch": 1793514882.5704014, "cache_hit_rate": 0.9217343122316111, "cache_misses": 1130, "stall_s": 1.341223921161145, "remote": {"remote_pairs": 14215, "local_pairs": 15737, "layers_remote": 3214, "remote_wait_s": 8.121419750386849, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 6520, "exec_wall_s": 20.257395823369734, "bytes_sent": 34036879, "bytes_recv": 133051428}}}}
27 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:35:14", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 1132.2989590698853, "prompt_tps": 15.068560844969106, "gen_tps": 4.587064207548293, "decode_tps_wall": 4.515398112320749, "itl_ms_p50": 222.81299997121096, "itl_ms_p95": 243.8272499712184, "itl_ms_p99": 248.51025000680238, "itl_ms_mean": 221.4644146817071, "total_s": 15.084557084017433, "peak_mem_gb": 5.237140036, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793514899.5282571, "t_end_epoch": 1793514914.617027, "t_first_epoch": 1793514900.660556, "cache_hit_rate": 0.9118762323676627, "cache_misses": 1162, "stall_s": 1.4608960741898045, "remote": {"remote_pairs": 16439, "local_pairs": 14665, "layers_remote": 3241, "remote_wait_s": 6.063985421205871, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 9761, "exec_wall_s": 27.9671216274146, "bytes_sent": 50382350, "bytes_recv": 200708431}}}}
28 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:35:32", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1109.9436670774594, "prompt_tps": 14.465983691319451, "gen_tps": 3.7254427743164378, "decode_tps_wall": 3.6672399624934817, "itl_ms_p50": 274.5515829883516, "itl_ms_p95": 289.32337497826666, "itl_ms_p99": 300.5584171041846, "itl_ms_mean": 272.68463755506906, "total_s": 18.28907583304681, "peak_mem_gb": 5.166995924, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793514914.617755, "t_end_epoch": 1793514932.909889, "t_first_epoch": 1793514915.7276986, "cache_hit_rate": 0.9306595113684508, "cache_misses": 982, "stall_s": 1.1886739959008992, "remote": {"remote_pairs": 14811, "local_pairs": 15909, "layers_remote": 3212, "remote_wait_s": 8.327266146894544, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 12973, "exec_wall_s": 37.66970552224666, "bytes_sent": 66508772, "bytes_recv": 261693858}}}}
29 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:35:50", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1354.9439159687608, "prompt_tps": 11.834137900138042, "gen_tps": 3.983599210826353, "decode_tps_wall": 3.9213707179501123, "itl_ms_p50": 265.87666606064886, "itl_ms_p95": 282.28720801416785, "itl_ms_p99": 291.7503329226747, "itl_ms_mean": 255.01286971478885, "total_s": 17.42075470800046, "peak_mem_gb": 5.144286508, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793514932.910409, "t_end_epoch": 1793514950.333293, "t_first_epoch": 1793514934.265353, "cache_hit_rate": 0.9508958195089582, "cache_misses": 740, "stall_s": 1.0301309142960235, "remote": {"remote_pairs": 14035, "local_pairs": 16685, "layers_remote": 3213, "remote_wait_s": 8.217726533068344, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 16186, "exec_wall_s": 47.087310496135615, "bytes_sent": 82601688, "bytes_recv": 319500650}}}}
30 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:36:03", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 1180.5846249917522, "prompt_tps": 12.728454545189736, "gen_tps": 5.1753948562283565, "decode_tps_wall": 5.094539833770586, "itl_ms_p50": 190.8072920050472, "itl_ms_p95": 227.61095804162323, "itl_ms_p99": 255.98308304324746, "itl_ms_mean": 196.2885820170096, "total_s": 13.546765292063355, "peak_mem_gb": 5.14983104, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793514950.334547, "t_end_epoch": 1793514963.883127, "t_first_epoch": 1793514951.5151317, "cache_hit_rate": 0.9501943960316396, "cache_misses": 743, "stall_s": 0.5246965067926794, "remote": {"remote_pairs": 13950, "local_pairs": 16386, "layers_remote": 3209, "remote_wait_s": 7.466171499807388, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 19395, "exec_wall_s": 55.13774217583705, "bytes_sent": 98487967, "bytes_recv": 376958852}}}}
31 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:36:16", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 1285.9910000115633, "prompt_tps": 14.00576298254918, "gen_tps": 5.468497560972256, "decode_tps_wall": 5.38306493540145, "itl_ms_p50": 183.5795419756323, "itl_ms_p95": 212.57733402308077, "itl_ms_p99": 233.7069580098614, "itl_ms_mean": 185.76777579321984, "total_s": 12.989360874984413, "peak_mem_gb": 5.246581892, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793514963.88332, "t_end_epoch": 1793514976.874177, "t_first_epoch": 1793514965.169311, "cache_hit_rate": 0.9140489642184557, "cache_misses": 1141, "stall_s": 0.7788582415087149, "remote": {"remote_pairs": 16744, "local_pairs": 14744, "layers_remote": 3246, "remote_wait_s": 6.665186170488596, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 22641, "exec_wall_s": 62.6296579411719, "bytes_sent": 115011905, "bytes_recv": 445864981}}}}
32 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:36:29", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1076.916916994378, "prompt_tps": 14.870607130345261, "gen_tps": 5.638331142790545, "decode_tps_wall": 5.550241427960826, "itl_ms_p50": 179.9311670474708, "itl_ms_p95": 197.6273750187829, "itl_ms_p99": 212.97129103913903, "itl_ms_mean": 180.1723425871589, "total_s": 12.42777449998539, "peak_mem_gb": 5.243179872, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793514976.874326, "t_end_epoch": 1793514989.3035252, "t_first_epoch": 1793514977.951243, "cache_hit_rate": 0.9121028744326778, "cache_misses": 1162, "stall_s": 0.7396673449547961, "remote": {"remote_pairs": 16201, "local_pairs": 14519, "layers_remote": 3231, "remote_wait_s": 6.310364342527464, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 25872, "exec_wall_s": 69.73004648415372, "bytes_sent": 131141322, "bytes_recv": 512544505}}}}
33 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:36:42", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 1091.7903749505058, "prompt_tps": 12.830787531105516, "gen_tps": 5.34778121857701, "decode_tps_wall": 5.264230916052405, "itl_ms_p50": 187.44320806581527, "itl_ms_p95": 219.74204201251268, "itl_ms_p99": 236.25983309466392, "itl_ms_mean": 189.96127182617784, "total_s": 13.05935049999971, "peak_mem_gb": 5.163703192, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793514989.3036811, "t_end_epoch": 1793515002.3644748, "t_first_epoch": 1793514990.3954716, "cache_hit_rate": 0.9168167336196149, "cache_misses": 1201, "stall_s": 0.7567394201178104, "remote": {"remote_pairs": 14215, "local_pairs": 15737, "layers_remote": 3214, "remote_wait_s": 6.478410501498729, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 29086, "exec_wall_s": 77.02638487343211, "bytes_sent": 146854089, "bytes_recv": 571087321}}}}
34 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:36:55", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1095.375958015211, "prompt_tps": 14.615925442823272, "gen_tps": 5.54208646933428, "decode_tps_wall": 5.455505481838607, "itl_ms_p50": 179.60200004745275, "itl_ms_p95": 210.48866701312363, "itl_ms_p99": 220.66566697321832, "itl_ms_mean": 183.30107142757032, "total_s": 12.643343457952142, "peak_mem_gb": 5.15038586, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793515002.3646169, "t_end_epoch": 1793515015.0102222, "t_first_epoch": 1793515003.459993, "cache_hit_rate": 0.9471798274717983, "cache_misses": 796, "stall_s": 0.5547539466060698, "remote": {"remote_pairs": 14035, "local_pairs": 16685, "layers_remote": 3213, "remote_wait_s": 6.448199286474846, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 32299, "exec_wall_s": 84.08878421050031, "bytes_sent": 162947005, "bytes_recv": 628892869}}}}
35 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:37:08", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1392.223917064257, "prompt_tps": 15.82075848096864, "gen_tps": 5.290335014798997, "decode_tps_wall": 5.207682516810356, "itl_ms_p50": 186.87750003300607, "itl_ms_p95": 220.38287506438792, "itl_ms_p99": 267.9635420208797, "itl_ms_mean": 192.02399469860313, "total_s": 13.489735583076254, "peak_mem_gb": 5.25269258, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515015.010528, "t_end_epoch": 1793515028.502267, "t_first_epoch": 1793515016.402752, "cache_hit_rate": 0.9134518133217743, "cache_misses": 1198, "stall_s": 0.7985853815916926, "remote": {"remote_pairs": 17174, "local_pairs": 15850, "layers_remote": 3257, "remote_wait_s": 6.68526480323635, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 35556, "exec_wall_s": 91.61451610759832, "bytes_sent": 180266083, "bytes_recv": 699560063}}}}
36 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:37:22", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 1237.7295419573784, "prompt_tps": 13.773447648697097, "gen_tps": 5.204947783774052, "decode_tps_wall": 5.1236360488961346, "itl_ms_p50": 192.74383294396102, "itl_ms_p95": 225.85683304350823, "itl_ms_p99": 229.2112500872463, "itl_ms_mean": 195.17389417529876, "total_s": 13.5336848750012, "peak_mem_gb": 5.21024854, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515028.502557, "t_end_epoch": 1793515042.038521, "t_first_epoch": 1793515029.7402866, "cache_hit_rate": 0.9213559836189898, "cache_misses": 1037, "stall_s": 0.729013200267218, "remote": {"remote_pairs": 16439, "local_pairs": 14665, "layers_remote": 3241, "remote_wait_s": 6.840626380755566, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 38797, "exec_wall_s": 99.2281679547159, "bytes_sent": 196611554, "bytes_recv": 767215123}}}}
37 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:37:35", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 1103.294291999191, "prompt_tps": 13.6156467126173, "gen_tps": 5.264907650464, "decode_tps_wall": 5.182652261805236, "itl_ms_p50": 186.99504202231765, "itl_ms_p95": 231.12516710534692, "itl_ms_p99": 242.5232919631526, "itl_ms_mean": 192.95139814216998, "total_s": 13.2592323749559, "peak_mem_gb": 5.170147084, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793515042.038788, "t_end_epoch": 1793515055.300355, "t_first_epoch": 1793515043.1420825, "cache_hit_rate": 0.922777852259016, "cache_misses": 1152, "stall_s": 0.7333188205957413, "remote": {"remote_pairs": 13950, "local_pairs": 16386, "layers_remote": 3209, "remote_wait_s": 6.456808029557578, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 42006, "exec_wall_s": 106.48047318425961, "bytes_sent": 212497833, "bytes_recv": 824672060}}}}
38 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:37:48", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1095.717207994312, "prompt_tps": 14.624452058793569, "gen_tps": 5.164295192761926, "decode_tps_wall": 5.0836123440656165, "itl_ms_p50": 194.7881249943748, "itl_ms_p95": 224.20083300676197, "itl_ms_p99": 241.480499971658, "itl_ms_mean": 196.71051455513825, "total_s": 13.488479624968022, "peak_mem_gb": 5.143712692, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793515055.3007848, "t_end_epoch": 1793515068.792329, "t_first_epoch": 1793515056.396502, "cache_hit_rate": 0.9626465188532694, "cache_misses": 529, "stall_s": 0.43906823929864913, "remote": {"remote_pairs": 14811, "local_pairs": 15909, "layers_remote": 3212, "remote_wait_s": 7.023411262547597, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 45218, "exec_wall_s": 114.00440316356253, "bytes_sent": 228624255, "bytes_recv": 885655905}}}}
39 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:38:02", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 1238.3940840372816, "prompt_tps": 14.560759521819962, "gen_tps": 5.306709882972244, "decode_tps_wall": 5.2237984789939915, "itl_ms_p50": 193.1757089914754, "itl_ms_p95": 206.77979208994657, "itl_ms_p99": 223.19725004490465, "itl_ms_mean": 191.43158068237383, "total_s": 13.298583667026833, "peak_mem_gb": 5.258764464, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515068.792758, "t_end_epoch": 1793515082.093204, "t_first_epoch": 1793515070.031152, "cache_hit_rate": 0.9165348399246704, "cache_misses": 1108, "stall_s": 0.81066174863372, "remote": {"remote_pairs": 16744, "local_pairs": 14744, "layers_remote": 3246, "remote_wait_s": 6.782545867492445, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 48464, "exec_wall_s": 121.65622012002859, "bytes_sent": 245148193, "bytes_recv": 954560698}}}}
40 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:38:15", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1282.4580420274287, "prompt_tps": 12.484395723803932, "gen_tps": 5.1322146022063215, "decode_tps_wall": 5.052039312799578, "itl_ms_p50": 204.32104193605483, "itl_ms_p95": 252.78858398087323, "itl_ms_p99": 258.20762501098216, "itl_ms_mean": 197.93986904782258, "total_s": 13.752669792040251, "peak_mem_gb": 5.228557464, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793515082.093418, "t_end_epoch": 1793515095.847877, "t_first_epoch": 1793515083.375876, "cache_hit_rate": 0.9117246596066566, "cache_misses": 1167, "stall_s": 0.7533264862140641, "remote": {"remote_pairs": 16201, "local_pairs": 14519, "layers_remote": 3231, "remote_wait_s": 7.6973897714633495, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 51695, "exec_wall_s": 130.14457180013414, "bytes_sent": 261277610, "bytes_recv": 1021239921}}}}
41 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:38:27", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 1036.1990000819787, "prompt_tps": 16.416337635782412, "gen_tps": 6.2152979201451135, "decode_tps_wall": 6.118191440764121, "itl_ms_p50": 162.71420801058412, "itl_ms_p95": 170.58641591574997, "itl_ms_p99": 186.76341709215194, "itl_ms_mean": 163.4469940474937, "total_s": 11.333359625074081, "peak_mem_gb": 5.277194848, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515095.848022, "t_end_epoch": 1793515107.183014, "t_first_epoch": 1793515096.884221, "cache_hit_rate": 0.9215834976490217, "cache_misses": 1034, "stall_s": 0.5873145798686892, "remote": {"remote_pairs": 16439, "local_pairs": 14665, "layers_remote": 3241, "remote_wait_s": 7.185467571718618, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 54936, "exec_wall_s": 137.95890715310816, "bytes_sent": 277623081, "bytes_recv": 1088893796}}}}
42 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:38:38", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1151.1061249766499, "prompt_tps": 19.122466383900427, "gen_tps": 6.358968912073889, "decode_tps_wall": 6.259616941999142, "itl_ms_p50": 159.3386660097167, "itl_ms_p95": 168.69108309037983, "itl_ms_p99": 172.36404202412814, "itl_ms_mean": 159.7541845237304, "total_s": 11.215619749971665, "peak_mem_gb": 5.283909132, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515107.183151, "t_end_epoch": 1793515118.400306, "t_first_epoch": 1793515108.3342571, "cache_hit_rate": 0.9174974714636613, "cache_misses": 1142, "stall_s": 0.6033356779953465, "remote": {"remote_pairs": 17174, "local_pairs": 15850, "layers_remote": 3257, "remote_wait_s": 7.258979036239907, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 58193, "exec_wall_s": 145.85528097953647, "bytes_sent": 294942159, "bytes_recv": 1159559849}}}}
43 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:38:49", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 908.91883301083, "prompt_tps": 16.514713722236642, "gen_tps": 6.592160376116109, "decode_tps_wall": 6.489165612303665, "itl_ms_p50": 153.39287498500198, "itl_ms_p95": 161.14683402702212, "itl_ms_p99": 166.3596669677645, "itl_ms_mean": 154.1030172051655, "total_s": 10.617408916936256, "peak_mem_gb": 5.18135054, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793515118.40044, "t_end_epoch": 1793515129.019375, "t_first_epoch": 1793515119.3093588, "cache_hit_rate": 0.9220404880010725, "cache_misses": 1163, "stall_s": 0.564133596024476, "remote": {"remote_pairs": 13950, "local_pairs": 16386, "layers_remote": 3209, "remote_wait_s": 6.855135958874598, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 61402, "exec_wall_s": 153.31614454556257, "bytes_sent": 310828438, "bytes_recv": 1217015622}}}}
44 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:38:59", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 970.3374580712989, "prompt_tps": 16.499479094189212, "gen_tps": 6.470247858243638, "decode_tps_wall": 6.3691567550064, "itl_ms_p50": 157.22358308266848, "itl_ms_p95": 164.38837500754744, "itl_ms_p99": 174.02379191480577, "itl_ms_mean": 157.00665542796727, "total_s": 10.861756750033237, "peak_mem_gb": 5.25391146, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793515129.019511, "t_end_epoch": 1793515139.88273, "t_first_epoch": 1793515129.9898484, "cache_hit_rate": 0.9048411497730711, "cache_misses": 1258, "stall_s": 0.6706733646569774, "remote": {"remote_pairs": 16201, "local_pairs": 14519, "layers_remote": 3231, "remote_wait_s": 6.996762365219183, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 64633, "exec_wall_s": 161.02084106509574, "bytes_sent": 326957855, "bytes_recv": 1283693698}}}}
45 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:39:11", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 916.7149579152465, "prompt_tps": 17.46524768752689, "gen_tps": 5.787001432866444, "decode_tps_wall": 5.69657981413533, "itl_ms_p50": 154.9672920955345, "itl_ms_p95": 286.6178329568356, "itl_ms_p99": 317.80129100661725, "itl_ms_mean": 175.54392857247933, "total_s": 11.975982457981445, "peak_mem_gb": 5.203072, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793515139.882868, "t_end_epoch": 1793515151.8619719, "t_first_epoch": 1793515140.799583, "cache_hit_rate": 0.921765096217651, "cache_misses": 1179, "stall_s": 0.6905802942346781, "remote": {"remote_pairs": 14035, "local_pairs": 16685, "layers_remote": 3213, "remote_wait_s": 7.113299403339624, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 67846, "exec_wall_s": 168.8945738732582, "bytes_sent": 343050771, "bytes_recv": 1341498073}}}}
46 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:39:27", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 1429.0841670008376, "prompt_tps": 9.81642353313077, "gen_tps": 4.352004617482674, "decode_tps_wall": 4.284027340655223, "itl_ms_p50": 261.98533293791115, "itl_ms_p95": 287.7565410453826, "itl_ms_p99": 312.73095903452486, "itl_ms_mean": 233.4252142861101, "total_s": 16.134872667025775, "peak_mem_gb": 5.17706814, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793515151.86248, "t_end_epoch": 1793515167.999059, "t_first_epoch": 1793515153.291564, "cache_hit_rate": 0.9441750935032553, "cache_misses": 806, "stall_s": 1.0984888303792104, "remote": {"remote_pairs": 14215, "local_pairs": 15737, "layers_remote": 3214, "remote_wait_s": 7.833401922485791, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 71060, "exec_wall_s": 177.97276857413817, "bytes_sent": 358763538, "bytes_recv": 1400039804}}}}
47 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:39:39", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 1054.9803749890998, "prompt_tps": 17.07129860046644, "gen_tps": 6.356307168516515, "decode_tps_wall": 6.256996238006367, "itl_ms_p50": 159.20950006693602, "itl_ms_p95": 165.60004197526723, "itl_ms_p99": 183.466040994972, "itl_ms_mean": 159.8210965711919, "total_s": 11.12370945897419, "peak_mem_gb": 5.267684492, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515167.999177, "t_end_epoch": 1793515179.124506, "t_first_epoch": 1793515169.0541573, "cache_hit_rate": 0.9114124293785311, "cache_misses": 1176, "stall_s": 0.6481640125857666, "remote": {"remote_pairs": 16744, "local_pairs": 14744, "layers_remote": 3246, "remote_wait_s": 7.179119484499097, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 74306, "exec_wall_s": 185.84211858559866, "bytes_sent": 375287476, "bytes_recv": 1468943501}}}}
48 +{"backend": "titan-distributed", "capacity": 32, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.6501669995486736, 0.47087494749575853, 0.5507919704541564, 0.6362079875543714, 0.5395410116761923, 0.5173750687390566, 0.5742919165641069, 0.5961250280961394, 0.8217920549213886, 0.5074170185253024]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:39:51", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 905.530667048879, "prompt_tps": 17.681796740739582, "gen_tps": 5.637963089245952, "decode_tps_wall": 5.549868856995565, "itl_ms_p50": 157.6134170172736, "itl_ms_p95": 275.9347080718726, "itl_ms_p99": 281.2341669341549, "itl_ms_mean": 180.18443782496016, "total_s": 12.257150250021368, "peak_mem_gb": 5.1775193, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793515179.1246479, "t_end_epoch": 1793515191.385385, "t_first_epoch": 1793515180.0301785, "cache_hit_rate": 0.9259285411665019, "cache_misses": 1049, "stall_s": 0.6167279224609956, "remote": {"remote_pairs": 14811, "local_pairs": 15909, "layers_remote": 3212, "remote_wait_s": 7.289832539507188, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 77518, "exec_wall_s": 193.8199627540307, "bytes_sent": 391413898, "bytes_recv": 1529926273}}}}
49 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:40:17", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1322.872249991633, "prompt_tps": 16.651080460047698, "gen_tps": 4.033169860253922, "decode_tps_wall": 3.970178581354954, "itl_ms_p50": 255.77166699804366, "itl_ms_p95": 292.84841602202505, "itl_ms_p99": 295.45933299232274, "itl_ms_mean": 251.87783861821075, "total_s": 17.19117608293891, "peak_mem_gb": 5.34836638, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515199.928149, "t_end_epoch": 1793515217.12319, "t_first_epoch": 1793515201.2510211, "cache_hit_rate": 0.8699902248289345, "cache_misses": 2261, "stall_s": 1.9478051020996645, "remote": {"remote_pairs": 12178, "local_pairs": 20846, "layers_remote": 3134, "remote_wait_s": 7.018351086298935, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 3182, "exec_wall_s": 9.452564351609908, "bytes_sent": 17770870, "bytes_recv": 52674328}}}}
50 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:40:26", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 839.8712079506367, "prompt_tps": 16.750737762026947, "gen_tps": 7.317069371699538, "decode_tps_wall": 7.202758279397386, "itl_ms_p50": 133.7072909809649, "itl_ms_p95": 172.24791599437594, "itl_ms_p99": 236.56300001312047, "itl_ms_mean": 138.83570171449156, "total_s": 9.586520415963605, "peak_mem_gb": 5.273616084, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793515217.12404, "t_end_epoch": 1793515226.712058, "t_first_epoch": 1793515217.963911, "cache_hit_rate": 0.8446530267875976, "cache_misses": 2946, "stall_s": 1.6031821323558688, "remote": {"remote_pairs": 9165, "local_pairs": 20787, "layers_remote": 2914, "remote_wait_s": 4.4731044742511585, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 6096, "exec_wall_s": 15.366762977675535, "bytes_sent": 32170280, "bytes_recv": 90503490}}}}
51 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:40:37", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 917.4251250224188, "prompt_tps": 18.54274562534336, "gen_tps": 6.674227431031557, "decode_tps_wall": 6.569955159476409, "itl_ms_p50": 151.6019169939682, "itl_ms_p95": 161.12649999558926, "itl_ms_p99": 231.9577910238877, "itl_ms_mean": 152.2080403482837, "total_s": 10.506531666964293, "peak_mem_gb": 5.301369432, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515226.712193, "t_end_epoch": 1793515237.220376, "t_first_epoch": 1793515227.6296182, "cache_hit_rate": 0.8686176142697882, "cache_misses": 2357, "stall_s": 1.032423764001578, "remote": {"remote_pairs": 10640, "local_pairs": 20464, "layers_remote": 3093, "remote_wait_s": 6.162299585877918, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 9189, "exec_wall_s": 22.52994364721235, "bytes_sent": 47850239, "bytes_recv": 134393123}}}}
52 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:40:47", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 817.2880420461297, "prompt_tps": 19.592438485524266, "gen_tps": 6.9454274251258585, "decode_tps_wall": 6.83691971376702, "itl_ms_p50": 146.23054198455065, "itl_ms_p95": 168.73429191764444, "itl_ms_p99": 223.11520809307694, "itl_ms_mean": 146.26469841182586, "total_s": 10.03196404199116, "peak_mem_gb": 5.194753828, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793515237.220522, "t_end_epoch": 1793515247.2539892, "t_first_epoch": 1793515238.03781, "cache_hit_rate": 0.8832177395721402, "cache_misses": 2391, "stall_s": 1.101161427446641, "remote": {"remote_pairs": 7531, "local_pairs": 23189, "layers_remote": 2851, "remote_wait_s": 5.605619699577801, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 12040, "exec_wall_s": 29.12705947854556, "bytes_sent": 62399551, "bytes_recv": 165524216}}}}
53 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:40:55", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 771.540749934502, "prompt_tps": 20.75362353087068, "gen_tps": 8.745899206636892, "decode_tps_wall": 8.609258991854867, "itl_ms_p50": 116.2234169896692, "itl_ms_p95": 145.43329109437764, "itl_ms_p99": 169.99400011263788, "itl_ms_mean": 116.15401522315565, "total_s": 8.089243708993308, "peak_mem_gb": 5.279729348, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793515247.254121, "t_end_epoch": 1793515255.344842, "t_first_epoch": 1793515248.0256617, "cache_hit_rate": 0.8769179452968646, "cache_misses": 2952, "stall_s": 1.335449076257646, "remote": {"remote_pairs": 4159, "local_pairs": 26561, "layers_remote": 1930, "remote_wait_s": 3.8053903091931716, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 13970, "exec_wall_s": 33.80501418397762, "bytes_sent": 73048930, "bytes_recv": 182751746}}}}
54 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:41:03", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 763.2551670540124, "prompt_tps": 19.669931985626327, "gen_tps": 9.221306751407148, "decode_tps_wall": 9.077241217785996, "itl_ms_p50": 106.66770895477384, "itl_ms_p95": 152.2026249440387, "itl_ms_p99": 198.54579190723598, "itl_ms_mean": 110.16563028429768, "total_s": 7.703689874964766, "peak_mem_gb": 5.266612788, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793515255.344975, "t_end_epoch": 1793515263.050188, "t_first_epoch": 1793515256.10823, "cache_hit_rate": 0.864248789025929, "cache_misses": 3335, "stall_s": 1.5100324804661795, "remote": {"remote_pairs": 3641, "local_pairs": 26695, "layers_remote": 1607, "remote_wait_s": 3.2609797889599577, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 15577, "exec_wall_s": 37.96051531389821, "bytes_sent": 82148303, "bytes_recv": 197825402}}}}
55 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:41:13", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 962.8030420280993, "prompt_tps": 18.707406547730567, "gen_tps": 6.55817361053182, "decode_tps_wall": 6.45571011404165, "itl_ms_p50": 151.11233305651695, "itl_ms_p95": 174.1400839528069, "itl_ms_p99": 293.3268330525607, "itl_ms_mean": 154.90162698367226, "total_s": 10.721605541999452, "peak_mem_gb": 5.323344932, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515263.050313, "t_end_epoch": 1793515273.7735229, "t_first_epoch": 1793515264.0131161, "cache_hit_rate": 0.9011393911909755, "cache_misses": 1744, "stall_s": 0.7411287919385359, "remote": {"remote_pairs": 11180, "local_pairs": 20308, "layers_remote": 3147, "remote_wait_s": 6.611293665715493, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 18724, "exec_wall_s": 45.345470033586025, "bytes_sent": 98152909, "bytes_recv": 243932211}}}}
56 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:41:23", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 833.0501249292865, "prompt_tps": 19.219876228606363, "gen_tps": 6.964716415442343, "decode_tps_wall": 6.855902078760631, "itl_ms_p50": 147.6216659648344, "itl_ms_p95": 160.39512504357845, "itl_ms_p99": 216.32287499960512, "itl_ms_mean": 145.85972619095136, "total_s": 10.022212874959223, "peak_mem_gb": 5.278263508, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793515273.773657, "t_end_epoch": 1793515283.797459, "t_first_epoch": 1793515274.606707, "cache_hit_rate": 0.8915644404058761, "cache_misses": 2148, "stall_s": 0.9856102492194623, "remote": {"remote_pairs": 8608, "local_pairs": 22112, "layers_remote": 2886, "remote_wait_s": 5.819438310805708, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 21610, "exec_wall_s": 52.049706744845025, "bytes_sent": 112770799, "bytes_recv": 279478181}}}}
57 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:41:33", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 832.8117090277374, "prompt_tps": 16.822767106002594, "gen_tps": 6.882824087218425, "decode_tps_wall": 6.775289190596227, "itl_ms_p50": 148.84620800148696, "itl_ms_p95": 158.7276660138741, "itl_ms_p99": 192.24883301649243, "itl_ms_mean": 147.5951759207491, "total_s": 10.131307792034931, "peak_mem_gb": 5.26486658, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793515283.797598, "t_end_epoch": 1793515293.930519, "t_first_epoch": 1793515284.6304095, "cache_hit_rate": 0.8378506644167897, "cache_misses": 3075, "stall_s": 1.358319653547369, "remote": {"remote_pairs": 9165, "local_pairs": 20787, "layers_remote": 2914, "remote_wait_s": 5.50597918278072, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 24524, "exec_wall_s": 58.7669949993724, "bytes_sent": 127170209, "bytes_recv": 317308381}}}}
58 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:41:42", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 773.0878749862313, "prompt_tps": 20.711618862055126, "gen_tps": 8.768451117001941, "decode_tps_wall": 8.631458900106184, "itl_ms_p50": 116.03529204148799, "itl_ms_p95": 144.65454197488725, "itl_ms_p99": 148.46024999860674, "itl_ms_mean": 115.8552698417759, "total_s": 8.071969875018112, "peak_mem_gb": 5.27621146, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793515293.9306471, "t_end_epoch": 1793515302.003971, "t_first_epoch": 1793515294.703735, "cache_hit_rate": 0.8751250833889259, "cache_misses": 2995, "stall_s": 1.3411274675745517, "remote": {"remote_pairs": 4159, "local_pairs": 26561, "layers_remote": 1930, "remote_wait_s": 3.788624220411293, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 26454, "exec_wall_s": 63.431973515078425, "bytes_sent": 137819588, "bytes_recv": 334535853}}}}
59 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:41:52", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1062.774125020951, "prompt_tps": 20.71253473133572, "gen_tps": 6.714892388731153, "decode_tps_wall": 6.6099813259612965, "itl_ms_p50": 153.00012496300042, "itl_ms_p95": 161.0748330131173, "itl_ms_p99": 164.82437495142221, "itl_ms_mean": 151.28635781048428, "total_s": 10.59381466708146, "peak_mem_gb": 5.321279908, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515302.004106, "t_end_epoch": 1793515312.5994189, "t_first_epoch": 1793515303.0668802, "cache_hit_rate": 0.8778103616813294, "cache_misses": 2125, "stall_s": 0.9161654679337516, "remote": {"remote_pairs": 12178, "local_pairs": 20846, "layers_remote": 3134, "remote_wait_s": 6.420234822318889, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 29588, "exec_wall_s": 70.7480551735498, "bytes_sent": 154591938, "bytes_recv": 384729270}}}}
60 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:42:03", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 876.7207920318469, "prompt_tps": 19.404100085889993, "gen_tps": 6.626007768495566, "decode_tps_wall": 6.522483262711839, "itl_ms_p50": 151.95349999703467, "itl_ms_p95": 170.15020898543298, "itl_ms_p99": 221.61362506449223, "itl_ms_mean": 153.31583995268576, "total_s": 10.53561870905105, "peak_mem_gb": 5.313054148, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515312.5995588, "t_end_epoch": 1793515323.1368299, "t_first_epoch": 1793515313.4762797, "cache_hit_rate": 0.8811036789297659, "cache_misses": 2133, "stall_s": 0.984493081108667, "remote": {"remote_pairs": 10640, "local_pairs": 20464, "layers_remote": 3093, "remote_wait_s": 6.209274447057396, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 32681, "exec_wall_s": 77.91377376334276, "bytes_sent": 170271897, "bytes_recv": 428618891}}}}
61 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:42:10", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 790.0047920411453, "prompt_tps": 19.001794560397098, "gen_tps": 9.481613638185921, "decode_tps_wall": 9.333483475226641, "itl_ms_p50": 105.13091599568725, "itl_ms_p95": 147.9442500276491, "itl_ms_p99": 156.43608395475894, "itl_ms_mean": 107.14113360293032, "total_s": 7.5398962090257555, "peak_mem_gb": 5.271270164, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793515323.136971, "t_end_epoch": 1793515330.678314, "t_first_epoch": 1793515323.9269757, "cache_hit_rate": 0.8567183620303659, "cache_misses": 3520, "stall_s": 1.5308039153460413, "remote": {"remote_pairs": 3641, "local_pairs": 26695, "layers_remote": 1607, "remote_wait_s": 3.1206252708798274, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 34288, "exec_wall_s": 81.94104974507354, "bytes_sent": 179371270, "bytes_recv": 443692430}}}}
62 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:42:20", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 758.7360000470653, "prompt_tps": 21.105276749240083, "gen_tps": 7.034013180902856, "decode_tps_wall": 6.924115825807954, "itl_ms_p50": 143.30762496683747, "itl_ms_p95": 163.66241697687656, "itl_ms_p99": 175.33300002105534, "itl_ms_mean": 144.42277182492293, "total_s": 9.85737062501721, "peak_mem_gb": 5.185114404, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793515330.678451, "t_end_epoch": 1793515340.537312, "t_first_epoch": 1793515331.437187, "cache_hit_rate": 0.8980658396014457, "cache_misses": 2087, "stall_s": 0.9895991479279473, "remote": {"remote_pairs": 7531, "local_pairs": 23189, "layers_remote": 2851, "remote_wait_s": 5.681811818270944, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 37139, "exec_wall_s": 88.51396049489267, "bytes_sent": 193920582, "bytes_recv": 474823060}}}}
63 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:42:30", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 968.6428749701008, "prompt_tps": 18.594007261175307, "gen_tps": 6.764233602225119, "decode_tps_wall": 6.658553770611622, "itl_ms_p50": 150.86745808366686, "itl_ms_p95": 157.9572909977287, "itl_ms_p99": 161.62441694177687, "itl_ms_mean": 150.18276257130006, "total_s": 10.430156916962005, "peak_mem_gb": 5.31601514, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515340.537447, "t_end_epoch": 1793515350.969048, "t_first_epoch": 1793515341.50609, "cache_hit_rate": 0.9017629386089224, "cache_misses": 1733, "stall_s": 0.737925092340447, "remote": {"remote_pairs": 11180, "local_pairs": 20308, "layers_remote": 3147, "remote_wait_s": 6.480890296748839, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 40286, "exec_wall_s": 95.76389819628093, "bytes_sent": 209925188, "bytes_recv": 520929376}}}}
64 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:42:41", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 833.6876248940825, "prompt_tps": 19.204986581659178, "gen_tps": 6.929657359508866, "decode_tps_wall": 6.821389648975483, "itl_ms_p50": 148.1790830148384, "itl_ms_p95": 160.160333965905, "itl_ms_p99": 254.46708290837705, "itl_ms_mean": 146.59769511190316, "total_s": 10.069342416943982, "peak_mem_gb": 5.279317516, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793515350.969184, "t_end_epoch": 1793515361.040412, "t_first_epoch": 1793515351.8028715, "cache_hit_rate": 0.891513958301782, "cache_misses": 2149, "stall_s": 0.9936705688014627, "remote": {"remote_pairs": 8608, "local_pairs": 22112, "layers_remote": 2886, "remote_wait_s": 5.790050574345514, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 43172, "exec_wall_s": 102.44672035600524, "bytes_sent": 224543078, "bytes_recv": 556474859}}}}
65 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:42:51", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 884.5940419705585, "prompt_tps": 19.230711214053738, "gen_tps": 6.6312850481191346, "decode_tps_wall": 6.527679166389206, "itl_ms_p50": 151.49408404249698, "itl_ms_p95": 169.91037502884865, "itl_ms_p99": 234.17533305473626, "itl_ms_mean": 153.19380357248028, "total_s": 10.535803667036816, "peak_mem_gb": 5.324360668, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515361.04054, "t_end_epoch": 1793515371.577888, "t_first_epoch": 1793515361.925134, "cache_hit_rate": 0.8760869565217392, "cache_misses": 2223, "stall_s": 1.0105739393038675, "remote": {"remote_pairs": 10640, "local_pairs": 20464, "layers_remote": 3093, "remote_wait_s": 6.2010979185579345, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 46265, "exec_wall_s": 109.62522316572722, "bytes_sent": 240223037, "bytes_recv": 600363962}}}}
66 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:43:02", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 975.3483749227598, "prompt_tps": 22.570222567413076, "gen_tps": 6.630296840512456, "decode_tps_wall": 6.526706510822484, "itl_ms_p50": 154.60541704669595, "itl_ms_p95": 162.22245909739286, "itl_ms_p99": 178.72254108078778, "itl_ms_mean": 153.21663358721824, "total_s": 10.627996290917508, "peak_mem_gb": 5.322108144, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515371.578024, "t_end_epoch": 1793515382.207586, "t_first_epoch": 1793515372.5533724, "cache_hit_rate": 0.887758035765626, "cache_misses": 1952, "stall_s": 0.8726274911314249, "remote": {"remote_pairs": 12178, "local_pairs": 20846, "layers_remote": 3134, "remote_wait_s": 6.5201878601219505, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 49399, "exec_wall_s": 116.99954978376627, "bytes_sent": 256995387, "bytes_recv": 650556885}}}}
67 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:43:10", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 818.7717919936404, "prompt_tps": 18.333539330274714, "gen_tps": 9.110465641066382, "decode_tps_wall": 8.968133074136, "itl_ms_p50": 109.15087501052767, "itl_ms_p95": 151.10250003635883, "itl_ms_p99": 201.3696669600904, "itl_ms_mean": 111.50592790421335, "total_s": 7.843645249959081, "peak_mem_gb": 5.261046548, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793515382.2077272, "t_end_epoch": 1793515390.05285, "t_first_epoch": 1793515383.026499, "cache_hit_rate": 0.8566369520087923, "cache_misses": 3522, "stall_s": 1.5799321399535984, "remote": {"remote_pairs": 3641, "local_pairs": 26695, "layers_remote": 1607, "remote_wait_s": 3.1827116645872593, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 51006, "exec_wall_s": 121.1244248348521, "bytes_sent": 266094760, "bytes_recv": 665630303}}}}
68 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:43:20", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 861.9180000387132, "prompt_tps": 18.575942152029928, "gen_tps": 6.377441210496598, "decode_tps_wall": 6.2777962457935494, "itl_ms_p50": 148.90354103408754, "itl_ms_p95": 265.1398750022054, "itl_ms_p99": 287.3195420252159, "itl_ms_mean": 159.29156679305294, "total_s": 10.897286708001047, "peak_mem_gb": 5.276057112, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793515390.052978, "t_end_epoch": 1793515400.953154, "t_first_epoch": 1793515390.914896, "cache_hit_rate": 0.8805593417133626, "cache_misses": 2366, "stall_s": 1.1809058682993054, "remote": {"remote_pairs": 8608, "local_pairs": 22112, "layers_remote": 2886, "remote_wait_s": 5.856127868872136, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 53892, "exec_wall_s": 128.0709979714593, "bytes_sent": 280712650, "bytes_recv": 701175762}}}}
69 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:43:33", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1034.1127079445869, "prompt_tps": 15.516690651211638, "gen_tps": 5.570872169517152, "decode_tps_wall": 5.483853724667841, "itl_ms_p50": 185.40504202246666, "itl_ms_p95": 237.2875000583008, "itl_ms_p99": 257.95850006397814, "itl_ms_mean": 182.3535145552356, "total_s": 12.522384124924429, "peak_mem_gb": 5.229614408, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793515400.953652, "t_end_epoch": 1793515413.478591, "t_first_epoch": 1793515401.9877646, "cache_hit_rate": 0.8645763842561708, "cache_misses": 3248, "stall_s": 3.0061631947755814, "remote": {"remote_pairs": 4159, "local_pairs": 26561, "layers_remote": 1930, "remote_wait_s": 3.02585978328716, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 55822, "exec_wall_s": 133.10669951105956, "bytes_sent": 291362029, "bytes_recv": 718402948}}}}
70 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:43:49", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 919.2907080287114, "prompt_tps": 15.27516020056432, "gen_tps": 4.189392792938808, "decode_tps_wall": 4.123942157613931, "itl_ms_p50": 245.0372080784291, "itl_ms_p95": 276.00779104977846, "itl_ms_p99": 295.2927079750225, "itl_ms_mean": 242.48642725352613, "total_s": 16.195935625000857, "peak_mem_gb": 5.19099108, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793515413.4790988, "t_end_epoch": 1793515429.678525, "t_first_epoch": 1793515414.3983896, "cache_hit_rate": 0.8549356675806792, "cache_misses": 2751, "stall_s": 3.024003808386624, "remote": {"remote_pairs": 9165, "local_pairs": 20787, "layers_remote": 2914, "remote_wait_s": 5.257889617350884, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 58736, "exec_wall_s": 141.15427585109137, "bytes_sent": 305761439, "bytes_recv": 756232670}}}}
71 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:44:07", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 1153.1106249894947, "prompt_tps": 15.66210630368204, "gen_tps": 3.823496450949806, "decode_tps_wall": 3.7637669859016145, "itl_ms_p50": 264.17287497315556, "itl_ms_p95": 287.9515419481322, "itl_ms_p99": 295.99308292381465, "itl_ms_mean": 265.69126190484633, "total_s": 17.891660124994814, "peak_mem_gb": 5.306900856, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515429.679181, "t_end_epoch": 1793515447.574995, "t_first_epoch": 1793515430.8322918, "cache_hit_rate": 0.8999489824839861, "cache_misses": 1765, "stall_s": 1.516195878968574, "remote": {"remote_pairs": 11180, "local_pairs": 20308, "layers_remote": 3147, "remote_wait_s": 7.6934304245514795, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 61883, "exec_wall_s": 150.52146257541608, "bytes_sent": 321766045, "bytes_recv": 802338968}}}}
72 +{"backend": "titan-distributed", "capacity": 32, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.806833035312593, 0.5772499134764075, 0.5125829484313726, 0.41562505066394806, 0.41212490759789944, 0.4094169707968831, 0.5832079332321882, 0.41045795660465956, 0.5816660122945905, 0.4081249935552478]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 32 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:44:22", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 916.2397080799565, "prompt_tps": 17.53842537050408, "gen_tps": 4.447769422339146, "decode_tps_wall": 4.378273506821364, "itl_ms_p50": 230.7572909630835, "itl_ms_p95": 259.61850001476705, "itl_ms_p99": 264.73266701214015, "itl_ms_mean": 228.400532411234, "total_s": 15.305473249987699, "peak_mem_gb": 5.190682956, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793515447.575671, "t_end_epoch": 1793515462.885622, "t_first_epoch": 1793515448.4919107, "cache_hit_rate": 0.8811663573312494, "cache_misses": 2433, "stall_s": 2.423531154752709, "remote": {"remote_pairs": 7531, "local_pairs": 23189, "layers_remote": 2851, "remote_wait_s": 5.311237015761435, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 64734, "exec_wall_s": 158.1122846325161, "bytes_sent": 336315357, "bytes_recv": 833469604}}}}
73 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:44:35", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1139.9086660239846, "prompt_tps": 19.324635472690026, "gen_tps": 19.783751234564004, "decode_tps_wall": 19.475127032765357, "itl_ms_p50": 48.39879192877561, "itl_ms_p95": 80.91587503440678, "itl_ms_p99": 104.82216603122652, "itl_ms_mean": 51.34754696683515, "total_s": 4.374804124934599, "peak_mem_gb": 9.337094816, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515470.629857, "t_end_epoch": 1793515475.00652, "t_first_epoch": 1793515471.7697656, "cache_hit_rate": 0.8781578277779797, "cache_misses": 3352, "stall_s": 1.9566169192548841, "remote": null}
74 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:44:39", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 573.6109999706969, "prompt_tps": 24.461973099117532, "gen_tps": 18.26965938414009, "decode_tps_wall": 17.984254574005618, "itl_ms_p50": 47.67108301166445, "itl_ms_p95": 104.59737502969801, "itl_ms_p99": 113.74566692393273, "itl_ms_mean": 55.60419509660394, "total_s": 4.076675291056745, "peak_mem_gb": 9.119708316, "text": "The Rideau Canal, located in Ontario, Canada, was constructed between 1826 and 1832 as a strategic military waterway to provide a secure alternative route between Montreal and Lake Ontario, bypassing ", "t_start_epoch": 1793515475.0068638, "t_end_epoch": 1793515479.08479, "t_first_epoch": 1793515475.5804749, "cache_hit_rate": 0.9171424334544376, "cache_misses": 2235, "stall_s": 1.3750380750279874, "remote": null}
75 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:44:43", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 469.0797079820186, "prompt_tps": 36.28634838320111, "gen_tps": 14.833437302667209, "decode_tps_wall": 14.601705313778611, "itl_ms_p50": 63.02908400539309, "itl_ms_p95": 124.50249993707985, "itl_ms_p99": 164.89245800767094, "itl_ms_mean": 68.4851514607934, "total_s": 4.783644250012003, "peak_mem_gb": 9.211541844, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515479.084918, "t_end_epoch": 1793515483.870917, "t_first_epoch": 1793515479.5539978, "cache_hit_rate": 0.9154841784915484, "cache_misses": 2305, "stall_s": 1.663128047133796, "remote": null}
76 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:44:47", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 559.4107090728357, "prompt_tps": 28.692877992620183, "gen_tps": 21.28880041321464, "decode_tps_wall": 20.956263405276353, "itl_ms_p50": 45.74566602241248, "itl_ms_p95": 87.49616693239659, "itl_ms_p99": 109.05341699253768, "itl_ms_mean": 47.71843055514471, "total_s": 3.5656718340469524, "peak_mem_gb": 9.11658568, "text": "Include a sentence about the beauty of music and a sentence about perseverance.\n\n**Dear Beginner,**\n\nStarting to learn the piano as an adult is one of the most beautiful gifts you can give yourself. M", "t_start_epoch": 1793515483.871339, "t_end_epoch": 1793515487.438003, "t_first_epoch": 1793515484.43075, "cache_hit_rate": 0.9324545725350015, "cache_misses": 1814, "stall_s": 1.073186130146496, "remote": null}
77 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:44:50", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 310.0123340263963, "prompt_tps": 51.71034655178786, "gen_tps": 26.37481151985402, "decode_tps_wall": 25.962804957858793, "itl_ms_p50": 36.20183398015797, "itl_ms_p95": 55.30637491028756, "itl_ms_p99": 82.53837504889816, "itl_ms_mean": 38.516639539646725, "total_s": 2.73656062502414, "peak_mem_gb": 9.09122276, "text": "The Mac Studio and MacBook Pro are both high-performance Apple devices designed for demanding workloads, including scientific computing. However, they differ significantly in form factor, performance,", "t_start_epoch": 1793515487.438137, "t_end_epoch": 1793515490.175786, "t_first_epoch": 1793515487.7481494, "cache_hit_rate": 0.9499852136625758, "cache_misses": 1353, "stall_s": 0.6775202032877132, "remote": null}
78 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:44:52", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 386.8228329811245, "prompt_tps": 38.83885635345025, "gen_tps": 27.64740109281534, "decode_tps_wall": 27.215585829389497, "itl_ms_p50": 36.278832936659455, "itl_ms_p95": 47.67591704148799, "itl_ms_p99": 56.22658389620483, "itl_ms_mean": 36.74365146019097, "total_s": 2.7016728749731556, "peak_mem_gb": 9.098006224, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety precautions should hikers take, and what are the benefits of hiking ", "t_start_epoch": 1793515490.175928, "t_end_epoch": 1793515492.8790588, "t_first_epoch": 1793515490.562751, "cache_hit_rate": 0.9533449078349525, "cache_misses": 1263, "stall_s": 0.6011423172894865, "remote": null}
79 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:44:58", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 473.5223330790177, "prompt_tps": 38.08193633789351, "gen_tps": 12.689290594185413, "decode_tps_wall": 12.491038795132303, "itl_ms_p50": 74.70779190771282, "itl_ms_p95": 120.27791701257229, "itl_ms_p99": 168.67670801002532, "itl_ms_mean": 80.05739285588442, "total_s": 5.517138082999736, "peak_mem_gb": 9.270750794, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515492.879262, "t_end_epoch": 1793515498.3981671, "t_first_epoch": 1793515493.3527844, "cache_hit_rate": 0.9188240456340501, "cache_misses": 2220, "stall_s": 1.8894343199208379, "remote": null}
80 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:45:03", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 453.0690829269588, "prompt_tps": 35.48592105797609, "gen_tps": 12.47630943053506, "decode_tps_wall": 12.281376370546178, "itl_ms_p50": 78.90137506183237, "itl_ms_p95": 141.76349993795156, "itl_ms_p99": 147.6052920334041, "itl_ms_mean": 81.42409855610735, "total_s": 5.582787291961722, "peak_mem_gb": 9.297299, "text": "import heapq\nfrom collections import defaultdict\nimport sys\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = {node: float('inf') for node in graph}\n distance", "t_start_epoch": 1793515498.3985522, "t_end_epoch": 1793515503.984123, "t_first_epoch": 1793515498.8516212, "cache_hit_rate": 0.9164063935329781, "cache_misses": 2275, "stall_s": 1.9449329456547275, "remote": null}
81 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:45:08", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 536.5610419539735, "prompt_tps": 26.19416959586934, "gen_tps": 16.435943188963574, "decode_tps_wall": 16.179176417606143, "itl_ms_p50": 53.24816599022597, "itl_ms_p95": 123.87933395802975, "itl_ms_p99": 135.04825008567423, "itl_ms_mean": 61.80784325411041, "total_s": 4.430455166962929, "peak_mem_gb": 9.122362536, "text": "The Rideau Canal, located in Ontario, Canada, was constructed between 1826 and 1832 as a strategic military waterway to provide a secure alternative route between Montreal and Lake Ontario, bypassing ", "t_start_epoch": 1793515503.984592, "t_end_epoch": 1793515508.416611, "t_first_epoch": 1793515504.521153, "cache_hit_rate": 0.916475124193668, "cache_misses": 2253, "stall_s": 1.4412548555992544, "remote": null}
82 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:45:11", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 415.8595419721678, "prompt_tps": 38.5494833686043, "gen_tps": 20.793521671987595, "decode_tps_wall": 20.468756453579942, "itl_ms_p50": 48.11895894818008, "itl_ms_p95": 76.77741593215615, "itl_ms_p99": 99.06329202931374, "itl_ms_mean": 48.85494642861424, "total_s": 3.493721166974865, "peak_mem_gb": 9.088666856, "text": "The Mac Studio and MacBook Pro are both high-performance Apple devices designed for demanding workloads, including scientific computing. However, they differ significantly in form factor, performance,", "t_start_epoch": 1793515508.416812, "t_end_epoch": 1793515511.9118578, "t_first_epoch": 1793515508.8326714, "cache_hit_rate": 0.9447360638769777, "cache_misses": 1495, "stall_s": 0.8944838367169723, "remote": null}
83 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:45:16", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 495.82895904313773, "prompt_tps": 44.446947387186356, "gen_tps": 15.434593536883463, "decode_tps_wall": 15.19344771148185, "itl_ms_p50": 59.54100005328655, "itl_ms_p95": 112.56562499329448, "itl_ms_p99": 149.26874998491257, "itl_ms_mean": 65.81784588920455, "total_s": 4.6423532500630245, "peak_mem_gb": 9.30600569, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515511.912051, "t_end_epoch": 1793515516.556694, "t_first_epoch": 1793515512.4078798, "cache_hit_rate": 0.9126531205699538, "cache_misses": 2403, "stall_s": 1.5714384583989158, "remote": null}
84 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:45:21", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 434.4761249376461, "prompt_tps": 39.284320558058944, "gen_tps": 15.76330429562217, "decode_tps_wall": 15.517041522597081, "itl_ms_p50": 53.70387504808605, "itl_ms_p95": 121.20066594798118, "itl_ms_p99": 173.47270902246237, "itl_ms_mean": 64.44527447733674, "total_s": 4.49452841700986, "peak_mem_gb": 9.227297576, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515516.5570822, "t_end_epoch": 1793515521.0528388, "t_first_epoch": 1793515516.9915583, "cache_hit_rate": 0.9247607523924761, "cache_misses": 2052, "stall_s": 1.449254703358747, "remote": null}
85 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:45:24", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 568.706791033037, "prompt_tps": 26.403337382175753, "gen_tps": 21.100282179612027, "decode_tps_wall": 20.770715813567733, "itl_ms_p50": 43.919999967329204, "itl_ms_p95": 74.11491603124887, "itl_ms_p99": 104.02916697785258, "itl_ms_mean": 48.144705698914116, "total_s": 3.6018232500646263, "peak_mem_gb": 9.137819344, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety precautions should hikers take, and what are the benefits of hiking ", "t_start_epoch": 1793515521.0529711, "t_end_epoch": 1793515524.6564338, "t_first_epoch": 1793515521.6216779, "cache_hit_rate": 0.9229064312363784, "cache_misses": 2087, "stall_s": 1.0191334170522168, "remote": null}
86 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:45:27", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 301.6480830265209, "prompt_tps": 53.17781764745541, "gen_tps": 25.075083223865004, "decode_tps_wall": 24.683451058775102, "itl_ms_p50": 39.61858304683119, "itl_ms_p95": 63.797625014558434, "itl_ms_p99": 68.55295901186764, "itl_ms_mean": 40.51297355539328, "total_s": 2.8539654170162976, "peak_mem_gb": 9.086520552, "text": "Include a sentence about the beauty of music and a sentence about perseverance.\n\n**Dear Beginner,**\n\nStarting to learn the piano as an adult is one of the most beautiful gifts you can give yourself. M", "t_start_epoch": 1793515524.656645, "t_end_epoch": 1793515527.511592, "t_first_epoch": 1793515524.9582932, "cache_hit_rate": 0.9634346142389038, "cache_misses": 982, "stall_s": 0.5923262159340084, "remote": null}
87 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:45:31", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 416.5846250252798, "prompt_tps": 43.26262271930406, "gen_tps": 16.29495013852646, "decode_tps_wall": 16.040348861081764, "itl_ms_p50": 58.37195797357708, "itl_ms_p95": 89.89699999801815, "itl_ms_p99": 96.75845794845372, "itl_ms_mean": 62.34278372998926, "total_s": 4.344180000014603, "peak_mem_gb": 9.268096586, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515527.511724, "t_end_epoch": 1793515531.858638, "t_first_epoch": 1793515527.9283087, "cache_hit_rate": 0.9216396080152114, "cache_misses": 2143, "stall_s": 1.389181197853759, "remote": null}
88 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:45:37", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 466.42912505194545, "prompt_tps": 34.500262820254086, "gen_tps": 12.255457739205054, "decode_tps_wall": 12.063981422024277, "itl_ms_p50": 82.30433310382068, "itl_ms_p95": 139.83620796352625, "itl_ms_p99": 151.34754194878042, "itl_ms_mean": 82.89137433305206, "total_s": 5.688585708034225, "peak_mem_gb": 9.299553592, "text": "import heapq\nfrom collections import defaultdict\nimport sys\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = {node: float('inf') for node in graph}\n distance", "t_start_epoch": 1793515531.859184, "t_end_epoch": 1793515537.54962, "t_first_epoch": 1793515532.3256133, "cache_hit_rate": 0.9161859268785596, "cache_misses": 2281, "stall_s": 1.9884109450504184, "remote": null}
89 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:45:42", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 436.4865829702467, "prompt_tps": 39.08379537196623, "gen_tps": 15.263256745721312, "decode_tps_wall": 15.024821362722264, "itl_ms_p50": 61.714916955679655, "itl_ms_p95": 107.03908302821219, "itl_ms_p99": 121.6326670255512, "itl_ms_mean": 66.55653174560044, "total_s": 4.629548082943074, "peak_mem_gb": 9.208767252, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515537.5499148, "t_end_epoch": 1793515542.181459, "t_first_epoch": 1793515537.9864013, "cache_hit_rate": 0.9282073845928207, "cache_misses": 1958, "stall_s": 1.427172260475345, "remote": null}
90 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:45:46", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 493.2829999597743, "prompt_tps": 44.74136823119959, "gen_tps": 15.862121156979748, "decode_tps_wall": 15.614295184494484, "itl_ms_p50": 52.6727499673143, "itl_ms_p95": 122.80837504658848, "itl_ms_p99": 152.89220900740474, "itl_ms_mean": 64.04387698479233, "total_s": 4.528047250001691, "peak_mem_gb": 9.327226042, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515542.181807, "t_end_epoch": 1793515546.712079, "t_first_epoch": 1793515542.67509, "cache_hit_rate": 0.9184689760459452, "cache_misses": 2243, "stall_s": 1.5280457076150924, "remote": null}
91 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:45:50", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 602.0079579902813, "prompt_tps": 25.005209420453482, "gen_tps": 21.08774226746278, "decode_tps_wall": 20.758340622241434, "itl_ms_p50": 46.50666704401374, "itl_ms_p95": 71.6285000089556, "itl_ms_p99": 107.33345802873373, "itl_ms_mean": 48.173407412370636, "total_s": 3.6369326249696314, "peak_mem_gb": 9.137819344, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety precautions should hikers take, and what are the benefits of hiking ", "t_start_epoch": 1793515546.71249, "t_end_epoch": 1793515550.350797, "t_first_epoch": 1793515547.314498, "cache_hit_rate": 0.9226478519448856, "cache_misses": 2094, "stall_s": 1.0354604698950425, "remote": null}
92 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:45:56", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 478.01616601645947, "prompt_tps": 33.522467516987135, "gen_tps": 11.084457816482894, "decode_tps_wall": 10.911274894751212, "itl_ms_p50": 88.97162496577948, "itl_ms_p95": 137.70195899996907, "itl_ms_p99": 163.10620901640505, "itl_ms_mean": 91.64831879371334, "total_s": 6.2518602500204, "peak_mem_gb": 9.27831992, "text": "import heapq\nfrom collections import defaultdict\nimport sys\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = {node: float('inf') for node in graph}\n distance", "t_start_epoch": 1793515550.3509781, "t_end_epoch": 1793515556.605464, "t_first_epoch": 1793515550.8289943, "cache_hit_rate": 0.9075142384714312, "cache_misses": 2517, "stall_s": 2.3027521276380867, "remote": null}
93 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:46:00", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 545.8548750029877, "prompt_tps": 29.44345430192379, "gen_tps": 17.68125735069977, "decode_tps_wall": 17.405033582569292, "itl_ms_p50": 54.47583307977766, "itl_ms_p95": 87.35145896207541, "itl_ms_p99": 112.24683304317296, "itl_ms_mean": 57.45464352343882, "total_s": 4.165497416979633, "peak_mem_gb": 9.11989476, "text": "The Mac Studio and MacBook Pro are both high-performance Apple devices designed for demanding workloads, including scientific computing. However, they differ significantly in form factor, performance,", "t_start_epoch": 1793515556.605917, "t_end_epoch": 1793515560.772984, "t_first_epoch": 1793515557.1517718, "cache_hit_rate": 0.920967026467544, "cache_misses": 2138, "stall_s": 1.2683848273009062, "remote": null}
94 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:46:05", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 442.15287489350885, "prompt_tps": 31.733825734244682, "gen_tps": 14.754844367996782, "decode_tps_wall": 14.524460657537208, "itl_ms_p50": 58.64229204598814, "itl_ms_p95": 132.0812499616295, "itl_ms_p99": 142.18620804604143, "itl_ms_mean": 68.8493723504334, "total_s": 4.779663332970813, "peak_mem_gb": 9.092697808, "text": "The Rideau Canal, located in Ontario, Canada, was constructed between 1826 and 1832 as a strategic military waterway to provide a secure alternative route between Montreal and Lake Ontario, bypassing ", "t_start_epoch": 1793515560.7732081, "t_end_epoch": 1793515565.554261, "t_first_epoch": 1793515561.215361, "cache_hit_rate": 0.9423148216801365, "cache_misses": 1556, "stall_s": 1.4374336481560022, "remote": null}
95 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:46:11", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 464.0621249563992, "prompt_tps": 38.84920034918969, "gen_tps": 12.359856151364571, "decode_tps_wall": 12.16673936924158, "itl_ms_p50": 78.96549999713898, "itl_ms_p95": 114.00579195469618, "itl_ms_p99": 160.60154105070978, "itl_ms_mean": 82.19128968341955, "total_s": 5.6421133750118315, "peak_mem_gb": 9.268145738, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515565.55442, "t_end_epoch": 1793515571.1987112, "t_first_epoch": 1793515566.0184822, "cache_hit_rate": 0.9169957583735556, "cache_misses": 2270, "stall_s": 1.8873011404648423, "remote": null}
96 +{"backend": "titan-distributed", "capacity": 64, "placement": "none", "workers": ["w0"], "rtt_ms": {"w0": [0.5463330307975411, 0.5532910581678152, 0.5979170091450214, 0.5696660373359919, 0.5845410050824285, 0.5723340436816216, 0.5835420452058315, 0.44620793778449297, 0.5770829739049077, 0.4640000406652689]}, "n_remote_experts": 0, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement none --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:46:15", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 568.7557499622926, "prompt_tps": 28.261034500639227, "gen_tps": 18.174762769699978, "decode_tps_wall": 17.890866354079055, "itl_ms_p50": 46.97899997700006, "itl_ms_p95": 109.08449999988079, "itl_ms_p99": 153.8282500114292, "itl_ms_mean": 55.894442460691884, "total_s": 4.090105624985881, "peak_mem_gb": 9.148436176, "text": "Include a sentence about the beauty of music and a sentence about perseverance.\n\n**Dear Beginner,**\n\nStarting to learn the piano as an adult is one of the most beautiful gifts you can give yourself. M", "t_start_epoch": 1793515571.1991858, "t_end_epoch": 1793515575.290367, "t_first_epoch": 1793515571.7679417, "cache_hit_rate": 0.9250819183795055, "cache_misses": 2012, "stall_s": 1.188437951495871, "remote": null}
97 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:46:38", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 963.7401250656694, "prompt_tps": 22.867216312868084, "gen_tps": 4.40785321726385, "decode_tps_wall": 4.339018438587864, "itl_ms_p50": 228.8271669531241, "itl_ms_p95": 258.600625093095, "itl_ms_p99": 303.29579196404666, "itl_ms_mean": 230.46687036560522, "total_s": 15.483152958098799, "peak_mem_gb": 9.26372558, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515582.7463338, "t_end_epoch": 1793515598.232702, "t_first_epoch": 1793515583.710074, "cache_hit_rate": 0.8949445878245028, "cache_misses": 1384, "stall_s": 1.3670447199838236, "remote": {"remote_pairs": 17174, "local_pairs": 15850, "layers_remote": 3165, "remote_wait_s": 7.060021840152331, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 3213, "exec_wall_s": 8.941859252634458, "bytes_sent": 18317194, "bytes_recv": 74520699}}}}
98 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:46:54", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 970.1113749761134, "prompt_tps": 14.486027966537664, "gen_tps": 4.165396991262915, "decode_tps_wall": 4.100324572244787, "itl_ms_p50": 248.02262498997152, "itl_ms_p95": 265.7001659972593, "itl_ms_p99": 280.75850009918213, "itl_ms_mean": 243.8831322693399, "total_s": 16.334748707944527, "peak_mem_gb": 9.163793744, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793515598.2334988, "t_end_epoch": 1793515614.5717041, "t_first_epoch": 1793515599.2036102, "cache_hit_rate": 0.973172987974098, "cache_misses": 377, "stall_s": 0.5553398623596877, "remote": {"remote_pairs": 14215, "local_pairs": 15737, "layers_remote": 3161, "remote_wait_s": 8.446504584047943, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 6374, "exec_wall_s": 18.123045922955498, "bytes_sent": 34025514, "bytes_recv": 133059817}}}}
99 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:47:11", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 1171.8634580029175, "prompt_tps": 14.556883646195747, "gen_tps": 4.11777833666132, "decode_tps_wall": 4.053459424862369, "itl_ms_p50": 248.76379198394716, "itl_ms_p95": 265.0392089271918, "itl_ms_p99": 275.88025003205985, "itl_ms_mean": 246.70285185695522, "total_s": 16.714143124991097, "peak_mem_gb": 9.163830636, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515614.572412, "t_end_epoch": 1793515631.290227, "t_first_epoch": 1793515615.7442756, "cache_hit_rate": 0.9884760112888052, "cache_misses": 147, "stall_s": 0.2969007956562564, "remote": {"remote_pairs": 16439, "local_pairs": 14665, "layers_remote": 3168, "remote_wait_s": 8.890568170812912, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 9542, "exec_wall_s": 27.470089736627415, "bytes_sent": 50365264, "bytes_recv": 200709066}}}}
100 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:47:27", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1084.7628329647705, "prompt_tps": 14.805213701952374, "gen_tps": 4.329426221378778, "decode_tps_wall": 4.261798901385038, "itl_ms_p50": 235.03016703762114, "itl_ms_p95": 261.80308393668383, "itl_ms_p99": 281.3988340785727, "itl_ms_mean": 234.64269974705067, "total_s": 15.867252917028964, "peak_mem_gb": 9.161102696, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793515631.2909162, "t_end_epoch": 1793515647.161278, "t_first_epoch": 1793515632.375679, "cache_hit_rate": 0.990599464892617, "cache_misses": 130, "stall_s": 0.19382941583171487, "remote": {"remote_pairs": 14811, "local_pairs": 15909, "layers_remote": 3165, "remote_wait_s": 8.734758967533708, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 12707, "exec_wall_s": 36.56761998415459, "bytes_sent": 66488060, "bytes_recv": 261689740}}}}
101 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:47:43", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1148.3987079700455, "prompt_tps": 13.981360662850696, "gen_tps": 4.103835074754196, "decode_tps_wall": 4.039722657001018, "itl_ms_p50": 248.16920794546604, "itl_ms_p95": 268.9505419693887, "itl_ms_p99": 283.37979095522314, "itl_ms_mean": 247.54174603223208, "total_s": 16.743528708000667, "peak_mem_gb": 9.161090388, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793515647.1619651, "t_end_epoch": 1793515663.907817, "t_first_epoch": 1793515648.3103638, "cache_hit_rate": 0.9963933310649881, "cache_misses": 53, "stall_s": 0.09953629516530782, "remote": {"remote_pairs": 14035, "local_pairs": 16685, "layers_remote": 3158, "remote_wait_s": 8.975231202319264, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 15865, "exec_wall_s": 45.81428337923717, "bytes_sent": 82576687, "bytes_recv": 319491296}}}}
102 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:48:00", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 1004.5342079829425, "prompt_tps": 14.973094592650618, "gen_tps": 4.201413865950144, "decode_tps_wall": 4.135777543957324, "itl_ms_p50": 244.6622090646997, "itl_ms_p95": 259.4484589062631, "itl_ms_p99": 268.92354106530547, "itl_ms_mean": 241.79250198335106, "total_s": 16.23746183293406, "peak_mem_gb": 9.161110884, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793515663.908318, "t_end_epoch": 1793515680.148764, "t_first_epoch": 1793515664.9128523, "cache_hit_rate": 0.9974685276409414, "cache_misses": 37, "stall_s": 0.07129120884928852, "remote": {"remote_pairs": 13950, "local_pairs": 16386, "layers_remote": 3156, "remote_wait_s": 8.799906125757843, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 19021, "exec_wall_s": 54.84993042878341, "bytes_sent": 98458720, "bytes_recv": 376944442}}}}
103 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:48:16", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 1143.2690829969943, "prompt_tps": 15.78013191530807, "gen_tps": 4.244754512932733, "decode_tps_wall": 4.178446008562962, "itl_ms_p50": 244.49937499593943, "itl_ms_p95": 264.88916599191725, "itl_ms_p99": 289.0197499655187, "itl_ms_mean": 239.32342261948165, "total_s": 16.220644708024338, "peak_mem_gb": 9.161110896, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515680.1492338, "t_end_epoch": 1793515696.3739572, "t_first_epoch": 1793515681.2925029, "cache_hit_rate": 0.9962320433315017, "cache_misses": 48, "stall_s": 0.10002329456619918, "remote": {"remote_pairs": 16744, "local_pairs": 14744, "layers_remote": 3164, "remote_wait_s": 8.985424359561875, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 22185, "exec_wall_s": 64.0774396955967, "bytes_sent": 114976256, "bytes_recv": 445842654}}}}
104 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:48:32", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1104.371124994941, "prompt_tps": 14.540843327463088, "gen_tps": 4.211221109650157, "decode_tps_wall": 4.145423370831197, "itl_ms_p50": 243.31929103937, "itl_ms_p95": 262.44554098229855, "itl_ms_p99": 284.8597909323871, "itl_ms_mean": 241.22988427102212, "total_s": 16.301853834069334, "peak_mem_gb": 9.161114976, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793515696.374629, "t_end_epoch": 1793515712.6796072, "t_first_epoch": 1793515697.479, "cache_hit_rate": 0.9978962131837307, "cache_misses": 27, "stall_s": 0.0627345428802073, "remote": {"remote_pairs": 16201, "local_pairs": 14519, "layers_remote": 3165, "remote_wait_s": 8.89705153775867, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 25350, "exec_wall_s": 73.18532142380718, "bytes_sent": 131100360, "bytes_recv": 512516826}}}}
105 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:48:49", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 1067.1643340028822, "prompt_tps": 13.166062175225063, "gen_tps": 4.173540509617236, "decode_tps_wall": 4.108340794213043, "itl_ms_p50": 245.80566701479256, "itl_ms_p95": 260.15741599258035, "itl_ms_p99": 279.840667033568, "itl_ms_mean": 243.40726587448327, "total_s": 16.40182208409533, "peak_mem_gb": 9.161110864, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793515712.6802628, "t_end_epoch": 1793515729.0853, "t_first_epoch": 1793515713.7474272, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 14215, "local_pairs": 15737, "layers_remote": 3161, "remote_wait_s": 9.009290341753513, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 28511, "exec_wall_s": 82.35021396377124, "bytes_sent": 146808680, "bytes_recv": 571055916}}}}
106 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:49:05", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1092.7697500446811, "prompt_tps": 14.689224206356513, "gen_tps": 4.132421765822029, "decode_tps_wall": 4.0678603148879375, "itl_ms_p50": 246.20779196266085, "itl_ms_p95": 260.4521659668535, "itl_ms_p99": 289.50429195538163, "itl_ms_mean": 245.82948346090103, "total_s": 16.580027208081447, "peak_mem_gb": 9.161090388, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793515729.085963, "t_end_epoch": 1793515745.669321, "t_first_epoch": 1793515730.1787329, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 14035, "local_pairs": 16685, "layers_remote": 3158, "remote_wait_s": 9.059909098199569, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 31669, "exec_wall_s": 91.5767612423515, "bytes_sent": 162897307, "bytes_recv": 628857426}}}}
107 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:49:22", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1253.8398751057684, "prompt_tps": 17.602733704457954, "gen_tps": 4.104942743484654, "decode_tps_wall": 4.040813974234167, "itl_ms_p50": 250.78995805233717, "itl_ms_p95": 264.80333297513425, "itl_ms_p99": 280.77808290254325, "itl_ms_mean": 247.47489153828826, "total_s": 16.84475804201793, "peak_mem_gb": 9.161110908, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515745.669973, "t_end_epoch": 1793515762.517947, "t_first_epoch": 1793515746.9238129, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 17174, "local_pairs": 15850, "layers_remote": 3165, "remote_wait_s": 9.309292294434272, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 34834, "exec_wall_s": 101.0324835695792, "bytes_sent": 180209561, "bytes_recv": 699517004}}}}
108 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:49:39", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 1114.1852079890668, "prompt_tps": 15.311645459556424, "gen_tps": 4.092627096674236, "decode_tps_wall": 4.028685594832943, "itl_ms_p50": 248.45616600941867, "itl_ms_p95": 259.34329198207706, "itl_ms_p99": 269.3942920304835, "itl_ms_mean": 248.2199160149321, "total_s": 16.75203991692979, "peak_mem_gb": 9.161110912, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515762.5186172, "t_end_epoch": 1793515779.273723, "t_first_epoch": 1793515763.6328022, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 16439, "local_pairs": 14665, "layers_remote": 3168, "remote_wait_s": 9.273902888176963, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 38002, "exec_wall_s": 110.4548593275249, "bytes_sent": 196549311, "bytes_recv": 767166328}}}}
109 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:49:55", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 1094.983167015016, "prompt_tps": 13.742450670424178, "gen_tps": 4.1970795081339425, "decode_tps_wall": 4.131512457291778, "itl_ms_p50": 246.23345804866403, "itl_ms_p95": 261.1775000113994, "itl_ms_p99": 271.3300839532167, "itl_ms_mean": 242.04211177799615, "total_s": 16.343636209028773, "peak_mem_gb": 9.161110884, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793515779.2743502, "t_end_epoch": 1793515795.6204329, "t_first_epoch": 1793515780.3693333, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 13950, "local_pairs": 16386, "layers_remote": 3156, "remote_wait_s": 9.009214854100719, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 41158, "exec_wall_s": 119.62753435887862, "bytes_sent": 212431344, "bytes_recv": 824619466}}}}
110 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:50:12", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1111.5457910345867, "prompt_tps": 14.432427318135066, "gen_tps": 4.151276964885927, "decode_tps_wall": 4.0864224842941095, "itl_ms_p50": 246.44070898648351, "itl_ms_p95": 263.48758302628994, "itl_ms_p99": 269.7420410113409, "itl_ms_mean": 244.7128273798005, "total_s": 16.528453915962018, "peak_mem_gb": 9.161111888, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793515795.621015, "t_end_epoch": 1793515812.153076, "t_first_epoch": 1793515796.7325609, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 14811, "local_pairs": 15909, "layers_remote": 3165, "remote_wait_s": 9.072284893831238, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 44323, "exec_wall_s": 128.85861845943145, "bytes_sent": 228554140, "bytes_recv": 885600165}}}}
111 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:50:28", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 1210.932124988176, "prompt_tps": 14.912571420620147, "gen_tps": 4.101889295778179, "decode_tps_wall": 4.037806775470187, "itl_ms_p50": 248.9270840305835, "itl_ms_p95": 267.1764160040766, "itl_ms_p99": 277.16199995484203, "itl_ms_mean": 247.6592010482111, "total_s": 16.813461791025475, "peak_mem_gb": 9.161114992, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515812.153753, "t_end_epoch": 1793515828.969561, "t_first_epoch": 1793515813.364685, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 16744, "local_pairs": 14744, "layers_remote": 3164, "remote_wait_s": 9.309203799581155, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 47487, "exec_wall_s": 138.311028562719, "bytes_sent": 245071676, "bytes_recv": 954498313}}}}
112 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:50:46", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1116.6006249841303, "prompt_tps": 14.367527953870567, "gen_tps": 4.0094462970941125, "decode_tps_wall": 3.946808321230094, "itl_ms_p50": 255.53412502631545, "itl_ms_p95": 278.5985000664368, "itl_ms_p99": 301.92100000567734, "itl_ms_mean": 253.36928439644416, "total_s": 17.078865541960113, "peak_mem_gb": 9.16111088, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793515828.9700649, "t_end_epoch": 1793515846.05131, "t_first_epoch": 1793515830.0866654, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 16201, "local_pairs": 14519, "layers_remote": 3165, "remote_wait_s": 9.391327627236024, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 50652, "exec_wall_s": 147.85371133650187, "bytes_sent": 261195780, "bytes_recv": 1021172439}}}}
113 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:51:03", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 1135.882708011195, "prompt_tps": 15.004187391932842, "gen_tps": 4.027945930918965, "decode_tps_wall": 3.9650194861120123, "itl_ms_p50": 252.5727499742061, "itl_ms_p95": 267.5045420182869, "itl_ms_p99": 277.9846249613911, "itl_ms_mean": 252.20557011198252, "total_s": 17.024833625066094, "peak_mem_gb": 9.161110892, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793515846.051817, "t_end_epoch": 1793515863.0797849, "t_first_epoch": 1793515847.1876996, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 16439, "local_pairs": 14665, "layers_remote": 3168, "remote_wait_s": 9.391926643555053, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 53820, "exec_wall_s": 157.39258473622613, "bytes_sent": 277535530, "bytes_recv": 1088821730}}}}
114 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:51:20", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1354.374916991219, "prompt_tps": 16.29116859588596, "gen_tps": 4.107028993604005, "decode_tps_wall": 4.04286812449914, "itl_ms_p50": 248.1480420101434, "itl_ms_p95": 266.79566595703363, "itl_ms_p99": 275.04129195585847, "itl_ms_mean": 247.34915144527184, "total_s": 16.937371458043344, "peak_mem_gb": 9.161110908, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515863.080438, "t_end_epoch": 1793515880.020965, "t_first_epoch": 1793515864.4348128, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 17174, "local_pairs": 15850, "layers_remote": 3165, "remote_wait_s": 9.3679064711323, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 56985, "exec_wall_s": 166.90655848523602, "bytes_sent": 294847784, "bytes_recv": 1159481324}}}}
115 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:51:36", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 1079.412917024456, "prompt_tps": 13.94455574927735, "gen_tps": 4.165509569205906, "decode_tps_wall": 4.100435069365417, "itl_ms_p50": 246.68387498240918, "itl_ms_p95": 263.8805409660563, "itl_ms_p99": 285.24491691496223, "itl_ms_mean": 243.87656019017516, "total_s": 16.44363620900549, "peak_mem_gb": 9.161110884, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793515880.021667, "t_end_epoch": 1793515896.4684072, "t_first_epoch": 1793515881.10108, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 13950, "local_pairs": 16386, "layers_remote": 3156, "remote_wait_s": 9.037361316033639, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 60141, "exec_wall_s": 176.10773200553376, "bytes_sent": 310729817, "bytes_recv": 1216934432}}}}
116 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:51:53", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1147.8684999747202, "prompt_tps": 13.982601864380449, "gen_tps": 4.127103158982259, "decode_tps_wall": 4.062633753543557, "itl_ms_p50": 248.294458957389, "itl_ms_p95": 268.92549998592585, "itl_ms_p99": 276.6487499466166, "itl_ms_mean": 246.14574206394278, "total_s": 16.655050250003114, "peak_mem_gb": 9.16111344, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793515896.469035, "t_end_epoch": 1793515913.1273742, "t_first_epoch": 1793515897.6169033, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 16201, "local_pairs": 14519, "layers_remote": 3165, "remote_wait_s": 9.206011266680434, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 63306, "exec_wall_s": 185.45556831732392, "bytes_sent": 326853921, "bytes_recv": 1283608553}}}}
117 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:52:09", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1105.9416249627247, "prompt_tps": 14.51760878265968, "gen_tps": 4.145329229987437, "decode_tps_wall": 4.080569572537088, "itl_ms_p50": 246.42770807258785, "itl_ms_p95": 263.6987909208983, "itl_ms_p99": 267.33079203404486, "itl_ms_mean": 245.06382803277424, "total_s": 16.5449627910275, "peak_mem_gb": 9.161090376, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793515913.128018, "t_end_epoch": 1793515929.676126, "t_first_epoch": 1793515914.2339594, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 14035, "local_pairs": 16685, "layers_remote": 3158, "remote_wait_s": 9.051426717080176, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 66464, "exec_wall_s": 194.67224440840073, "bytes_sent": 342942548, "bytes_recv": 1341409360}}}}
118 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:52:26", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 1079.5875419862568, "prompt_tps": 13.013700332317367, "gen_tps": 4.100192112784866, "decode_tps_wall": 4.036137385133188, "itl_ms_p50": 251.41945795621723, "itl_ms_p95": 268.01204099319875, "itl_ms_p99": 282.4765839613974, "itl_ms_mean": 247.76163558837854, "total_s": 16.688570584054105, "peak_mem_gb": 9.161110864, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793515929.676773, "t_end_epoch": 1793515946.368463, "t_first_epoch": 1793515930.7563605, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 14215, "local_pairs": 15737, "layers_remote": 3161, "remote_wait_s": 9.09321897814516, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 69625, "exec_wall_s": 203.92560902936384, "bytes_sent": 358650868, "bytes_recv": 1399947402}}}}
119 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:52:43", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 1146.2137500056997, "prompt_tps": 15.754968795966613, "gen_tps": 4.078027177153667, "decode_tps_wall": 4.014319555750858, "itl_ms_p50": 250.31191704329103, "itl_ms_p95": 273.75958301126957, "itl_ms_p99": 284.5238330774009, "itl_ms_mean": 249.10822023807597, "total_s": 16.840031625004485, "peak_mem_gb": 9.161114992, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793515946.369063, "t_end_epoch": 1793515963.212094, "t_first_epoch": 1793515947.5152767, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 16744, "local_pairs": 14744, "layers_remote": 3164, "remote_wait_s": 9.279039492597803, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 72789, "exec_wall_s": 213.35657692549285, "bytes_sent": 375168404, "bytes_recv": 1468844622}}}}
120 +{"backend": "titan-distributed", "capacity": 64, "placement": "odd", "workers": ["w0"], "rtt_ms": {"w0": [0.5886249709874392, 0.6431660149246454, 0.5612079985439777, 0.5914999637752771, 0.56245899759233, 0.5604999605566263, 0.554832979105413, 0.556000042706728, 0.5729160038754344, 0.38350000977516174]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement odd --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.653533935546875, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:53:00", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 1096.8299160012975, "prompt_tps": 14.626972765581936, "gen_tps": 4.003595019896459, "decode_tps_wall": 3.9410474044983106, "itl_ms_p50": 245.13520894106477, "itl_ms_p95": 328.6522919079289, "itl_ms_p99": 329.8474580515176, "itl_ms_mean": 253.73965277824374, "total_s": 17.082428041030653, "peak_mem_gb": 9.161110864, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793515963.212628, "t_end_epoch": 1793515980.2986088, "t_first_epoch": 1793515964.3094578, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 14811, "local_pairs": 15909, "layers_remote": 3165, "remote_wait_s": 9.105861554387957, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 75954, "exec_wall_s": 222.63520523917396, "bytes_sent": 391291200, "bytes_recv": 1529824295}}}}
121 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:53:24", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1115.895041031763, "prompt_tps": 19.743005885049367, "gen_tps": 4.0742582910258, "decode_tps_wall": 4.010639313451848, "itl_ms_p50": 252.01979198027402, "itl_ms_p95": 321.7072089901194, "itl_ms_p99": 337.6011250074953, "itl_ms_mean": 249.33680688910601, "total_s": 16.82411387504544, "peak_mem_gb": 9.29704792, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793515988.00973, "t_end_epoch": 1793516004.835907, "t_first_epoch": 1793515989.1256251, "cache_hit_rate": 0.9006237769080235, "cache_misses": 1625, "stall_s": 1.2717664138181135, "remote": {"remote_pairs": 12178, "local_pairs": 20846, "layers_remote": 3043, "remote_wait_s": 7.810057691880502, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 3091, "exec_wall_s": 9.652536986395717, "bytes_sent": 17767194, "bytes_recv": 52671783}}}}
122 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:53:40", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 716.8463330017403, "prompt_tps": 19.575239422004916, "gen_tps": 4.145246948579021, "decode_tps_wall": 4.080485382904077, "itl_ms_p50": 238.3082079468295, "itl_ms_p95": 307.60870908852667, "itl_ms_p99": 322.2315420862287, "itl_ms_mean": 245.0688842532506, "total_s": 16.156186040956527, "peak_mem_gb": 9.163847016, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793516004.836261, "t_end_epoch": 1793516020.995595, "t_first_epoch": 1793516005.5531073, "cache_hit_rate": 0.9778635918633743, "cache_misses": 407, "stall_s": 0.5163357534911484, "remote": {"remote_pairs": 9165, "local_pairs": 20787, "layers_remote": 2866, "remote_wait_s": 7.494642423815094, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 5957, "exec_wall_s": 17.835838578874245, "bytes_sent": 32203178, "bytes_recv": 90495816}}}}
123 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:53:56", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 1012.8243750659749, "prompt_tps": 16.8518546551645, "gen_tps": 4.3515464942738245, "decode_tps_wall": 4.283582110857139, "itl_ms_p50": 235.22062494885176, "itl_ms_p95": 264.95449990034103, "itl_ms_p99": 277.6893749833107, "itl_ms_mean": 233.4494761908279, "total_s": 15.720141375088133, "peak_mem_gb": 9.161112916, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793516020.9962938, "t_end_epoch": 1793516036.719563, "t_first_epoch": 1793516022.009118, "cache_hit_rate": 0.9972041006523765, "cache_misses": 48, "stall_s": 0.09839570487383753, "remote": {"remote_pairs": 10640, "local_pairs": 20464, "layers_remote": 3022, "remote_wait_s": 8.335200919071212, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 8979, "exec_wall_s": 26.45188994088676, "bytes_sent": 47889103, "bytes_recv": 134376899}}}}
124 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:54:11", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 902.7465829858556, "prompt_tps": 17.804154297313648, "gen_tps": 4.586478883348258, "decode_tps_wall": 4.514840953963704, "itl_ms_p50": 223.9991669775918, "itl_ms_p95": 240.9017919562757, "itl_ms_p99": 245.23858400061727, "itl_ms_mean": 221.49174471407952, "total_s": 14.856726499972865, "peak_mem_gb": 9.16111086, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793516036.7202501, "t_end_epoch": 1793516051.580273, "t_first_epoch": 1793516037.6229968, "cache_hit_rate": 0.9952787543947765, "cache_misses": 94, "stall_s": 0.18408004636876285, "remote": {"remote_pairs": 7531, "local_pairs": 23189, "layers_remote": 2805, "remote_wait_s": 7.480867761420086, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 11784, "exec_wall_s": 34.29861062660348, "bytes_sent": 62450755, "bytes_recv": 165502035}}}}
125 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:54:22", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 805.9088750742376, "prompt_tps": 19.95408273797211, "gen_tps": 6.125833522383629, "decode_tps_wall": 6.030142433094948, "itl_ms_p50": 162.1782920556143, "itl_ms_p95": 218.69283402338624, "itl_ms_p99": 236.97324993554503, "itl_ms_mean": 165.83356215795945, "total_s": 11.253423291025683, "peak_mem_gb": 9.16111344, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793516051.580937, "t_end_epoch": 1793516062.8372538, "t_first_epoch": 1793516052.3868458, "cache_hit_rate": 0.999058622165169, "cache_misses": 22, "stall_s": 0.04924375133123249, "remote": {"remote_pairs": 4159, "local_pairs": 26561, "layers_remote": 1877, "remote_wait_s": 4.8191085293656215, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 13661, "exec_wall_s": 39.29061028617434, "bytes_sent": 73099582, "bytes_recv": 182723330}}}}
126 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:54:32", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 746.8328750692308, "prompt_tps": 20.198209500229183, "gen_tps": 7.03524188999839, "decode_tps_wall": 6.925355250845889, "itl_ms_p50": 142.84641691483557, "itl_ms_p95": 186.94049993064255, "itl_ms_p99": 217.8760829847306, "itl_ms_mean": 144.39692460222258, "total_s": 9.843839125009254, "peak_mem_gb": 9.161110868, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793516062.8379369, "t_end_epoch": 1793516072.6848042, "t_first_epoch": 1793516063.5847697, "cache_hit_rate": 0.9990435795076513, "cache_misses": 23, "stall_s": 0.03993162699043751, "remote": {"remote_pairs": 3641, "local_pairs": 26695, "layers_remote": 1555, "remote_wait_s": 3.676505342242308, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 15216, "exec_wall_s": 43.11086757527664, "bytes_sent": 82198541, "bytes_recv": 197791023}}}}
127 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:54:48", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 1016.3911670679227, "prompt_tps": 17.779556470426506, "gen_tps": 4.247764551095592, "decode_tps_wall": 4.181406955841572, "itl_ms_p50": 240.79620908014476, "itl_ms_p95": 266.00537495687604, "itl_ms_p99": 273.6405419418588, "itl_ms_mean": 239.1539523803023, "total_s": 16.083090167026967, "peak_mem_gb": 9.161090376, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793516072.685445, "t_end_epoch": 1793516088.7705579, "t_first_epoch": 1793516073.7018363, "cache_hit_rate": 0.9994010900161706, "cache_misses": 10, "stall_s": 0.026632293011061847, "remote": {"remote_pairs": 11180, "local_pairs": 20308, "layers_remote": 3067, "remote_wait_s": 8.602350956643932, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 18283, "exec_wall_s": 51.92312859650701, "bytes_sent": 98269696, "bytes_recv": 243888430}}}}
128 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:55:03", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 879.7342920443043, "prompt_tps": 18.214794555483454, "gen_tps": 4.579652195614689, "decode_tps_wall": 4.508108490670101, "itl_ms_p50": 229.15420797653496, "itl_ms_p95": 253.03091597743332, "itl_ms_p99": 263.19387496914715, "itl_ms_mean": 221.82252314237374, "total_s": 14.85455325001385, "peak_mem_gb": 9.161102696, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793516088.770816, "t_end_epoch": 1793516103.6287959, "t_first_epoch": 1793516089.6505504, "cache_hit_rate": 0.9996343119841187, "cache_misses": 7, "stall_s": 0.01876679202541709, "remote": {"remote_pairs": 8608, "local_pairs": 22112, "layers_remote": 2824, "remote_wait_s": 7.770248188404366, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 21107, "exec_wall_s": 59.889764891471714, "bytes_sent": 112910689, "bytes_recv": 279426830}}}}
129 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:55:18", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 874.3489170446992, "prompt_tps": 16.052924198203264, "gen_tps": 4.552778655784819, "decode_tps_wall": 4.481658426191054, "itl_ms_p50": 221.5682080714032, "itl_ms_p95": 252.8453340055421, "itl_ms_p99": 258.92137503251433, "itl_ms_mean": 223.13168584110423, "total_s": 14.931645125034265, "peak_mem_gb": 9.161110892, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793516103.629228, "t_end_epoch": 1793516118.5636768, "t_first_epoch": 1793516104.503577, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 9165, "local_pairs": 20787, "layers_remote": 2866, "remote_wait_s": 7.875036829034798, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 23973, "exec_wall_s": 67.94024919334333, "bytes_sent": 127346673, "bytes_recv": 317250880}}}}
130 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:55:29", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 732.414250029251, "prompt_tps": 21.947996726098562, "gen_tps": 6.339016782787792, "decode_tps_wall": 6.2399906341271825, "itl_ms_p50": 155.11483396403491, "itl_ms_p95": 210.25112504139543, "itl_ms_p99": 234.15291705168784, "itl_ms_mean": 160.25665079221307, "total_s": 10.828583249938674, "peak_mem_gb": 9.16111344, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793516118.564112, "t_end_epoch": 1793516129.3957162, "t_first_epoch": 1793516119.2965262, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 4159, "local_pairs": 26561, "layers_remote": 1877, "remote_wait_s": 4.562393712811172, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 25850, "exec_wall_s": 72.63393346813973, "bytes_sent": 137995500, "bytes_recv": 334472174}}}}
131 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:55:45", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1064.6219160407782, "prompt_tps": 20.742126570068322, "gen_tps": 4.2925918708670086, "decode_tps_wall": 4.2255333840842155, "itl_ms_p50": 237.36229096539319, "itl_ms_p95": 264.8526669945568, "itl_ms_p99": 296.37687490321696, "itl_ms_mean": 236.65651388924627, "total_s": 15.973982291063294, "peak_mem_gb": 9.16110272, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793516129.396348, "t_end_epoch": 1793516145.373734, "t_first_epoch": 1793516130.46097, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 12178, "local_pairs": 20846, "layers_remote": 3043, "remote_wait_s": 8.596037587034516, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 28893, "exec_wall_s": 81.4040281282505, "bytes_sent": 154760078, "bytes_recv": 384654997}}}}
132 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:56:00", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 947.9092920664698, "prompt_tps": 18.008496839769613, "gen_tps": 4.453505369240884, "decode_tps_wall": 4.383926059181481, "itl_ms_p50": 227.34287497587502, "itl_ms_p95": 256.56408397480845, "itl_ms_p99": 262.20129104331136, "itl_ms_mean": 228.10603703172612, "total_s": 15.318589625065215, "peak_mem_gb": 9.161112936, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793516145.374397, "t_end_epoch": 1793516160.696683, "t_first_epoch": 1793516146.3223064, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 10640, "local_pairs": 20464, "layers_remote": 3022, "remote_wait_s": 8.215350259444676, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 31915, "exec_wall_s": 89.80087266908959, "bytes_sent": 170446003, "bytes_recv": 428536118}}}}
133 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:56:10", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 769.841791014187, "prompt_tps": 19.579406085164035, "gen_tps": 6.988413369600305, "decode_tps_wall": 6.879259410691379, "itl_ms_p50": 142.6802920177579, "itl_ms_p95": 192.96858308371156, "itl_ms_p99": 212.83004200085998, "itl_ms_mean": 145.3644847940831, "total_s": 9.927804333041422, "peak_mem_gb": 9.161110868, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793516160.6973891, "t_end_epoch": 1793516170.627423, "t_first_epoch": 1793516161.467231, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 3641, "local_pairs": 26695, "layers_remote": 1555, "remote_wait_s": 3.7370952035998926, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 33470, "exec_wall_s": 93.65224335866515, "bytes_sent": 179544962, "bytes_recv": 443603874}}}}
134 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:56:25", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 846.1375420447439, "prompt_tps": 18.97480605222471, "gen_tps": 4.703402619475945, "decode_tps_wall": 4.629928611825463, "itl_ms_p50": 217.95962506439537, "itl_ms_p95": 241.8932500295341, "itl_ms_p99": 258.11987498309463, "itl_ms_mean": 215.98605158746184, "total_s": 14.45325879205484, "peak_mem_gb": 9.16111086, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793516170.627925, "t_end_epoch": 1793516185.084215, "t_first_epoch": 1793516171.4740624, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 7531, "local_pairs": 23189, "layers_remote": 2805, "remote_wait_s": 7.483184522017837, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 36275, "exec_wall_s": 101.32391842734069, "bytes_sent": 194106614, "bytes_recv": 474729056}}}}
135 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:56:40", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 963.961667031981, "prompt_tps": 18.75104742196582, "gen_tps": 4.363015613125885, "decode_tps_wall": 4.294859255884006, "itl_ms_p50": 231.17237503174692, "itl_ms_p95": 256.3417919445783, "itl_ms_p99": 284.2176250414923, "itl_ms_mean": 232.8365006676269, "total_s": 15.632661209092475, "peak_mem_gb": 9.161090408, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793516185.0848799, "t_end_epoch": 1793516200.72066, "t_first_epoch": 1793516186.0488415, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 11180, "local_pairs": 20308, "layers_remote": 3067, "remote_wait_s": 8.451475383364595, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 39342, "exec_wall_s": 109.95624039613176, "bytes_sent": 210177769, "bytes_recv": 520826472}}}}
136 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:56:55", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 941.1168330116197, "prompt_tps": 17.0720417097501, "gen_tps": 4.674294376515088, "decode_tps_wall": 4.601276744267862, "itl_ms_p50": 221.9010000117123, "itl_ms_p95": 249.0918330149725, "itl_ms_p99": 264.11112502682954, "itl_ms_mean": 217.33098345926948, "total_s": 14.632968790945597, "peak_mem_gb": 9.161110888, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793516200.7213051, "t_end_epoch": 1793516215.3581178, "t_first_epoch": 1793516201.662422, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 8608, "local_pairs": 22112, "layers_remote": 2824, "remote_wait_s": 7.6650262408657, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 42166, "exec_wall_s": 117.80076284543611, "bytes_sent": 224818762, "bytes_recv": 556364855}}}}
137 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a bash script that lists the ten largest files under a", "ts": "2026-11-01T01:57:10", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 998.9557090448216, "prompt_tps": 17.083237373466343, "gen_tps": 4.447585577395119, "decode_tps_wall": 4.378100508473243, "itl_ms_p50": 228.0160840600729, "itl_ms_p95": 262.9119580378756, "itl_ms_p99": 273.57637509703636, "itl_ms_mean": 228.40955753862443, "total_s": 15.388757833978161, "peak_mem_gb": 9.161110868, "text": "#!/bin/bash\n\n# Check if directory is provided as argument\nif [ $# -eq 0 ]; then\n echo \"Usage: $0 <directory>\"\n exit 1\nfi\n\nDIR=\"$1\"\n\n# Check if directory exists\nif [ ! -d \"$DIR\" ]; then\n echo ", "t_start_epoch": 1793516215.358755, "t_end_epoch": 1793516230.7512808, "t_first_epoch": 1793516216.3577108, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 10640, "local_pairs": 20464, "layers_remote": 3022, "remote_wait_s": 8.242730411468074, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 45188, "exec_wall_s": 126.22563437605277, "bytes_sent": 240504687, "bytes_recv": 600246037}}}}
138 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a C function that computes the dot product of two floa", "ts": "2026-11-01T01:57:26", "prompt_tokens": 22, "gen_tokens": 64, "ttft_ms": 1071.2234999518842, "prompt_tps": 20.61412201896506, "gen_tps": 4.343152710344352, "decode_tps_wall": 4.275299109439543, "itl_ms_p50": 237.39995900541544, "itl_ms_p95": 259.1489580227062, "itl_ms_p99": 263.5341249406338, "itl_ms_mean": 233.90176322215075, "total_s": 15.80703458294738, "peak_mem_gb": 9.16110272, "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float* a, const float* b, int n) {\n float32x4_t sum_vec = vdupq_n_f32(0.0f);\n int i = 0;\n\n // Process 4 elements at a", "t_start_epoch": 1793516230.751945, "t_end_epoch": 1793516246.562152, "t_first_epoch": 1793516231.8231685, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 12178, "local_pairs": 20846, "layers_remote": 3043, "remote_wait_s": 8.538449062616564, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 48231, "exec_wall_s": 134.93780728487764, "bytes_sent": 257269265, "bytes_recv": 650428867}}}}
139 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "general", "prompt": "What should a visitor know before hiking in Gatineau Park in", "ts": "2026-11-01T01:57:36", "prompt_tokens": 15, "gen_tokens": 64, "ttft_ms": 741.6354579618201, "prompt_tps": 20.33414305575404, "gen_tps": 7.080746085168922, "decode_tps_wall": 6.9701566932760155, "itl_ms_p50": 138.94204201642424, "itl_ms_p95": 193.6276670312509, "itl_ms_p99": 210.03600000403821, "itl_ms_mean": 143.46879761895195, "total_s": 9.780169707955793, "peak_mem_gb": 9.161110868, "text": "What are the most popular winter hiking trails, and what should hikers expect in terms of weather and trail conditions? What safety tips should hikers follow, and what gear is essential for a safe and", "t_start_epoch": 1793516246.5628102, "t_end_epoch": 1793516256.346077, "t_first_epoch": 1793516247.3044457, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 3641, "local_pairs": 26695, "layers_remote": 1555, "remote_wait_s": 3.6209025020943955, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 49786, "exec_wall_s": 138.67085399769712, "bytes_sent": 266368224, "bytes_recv": 665496594}}}}
140 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement Dijkstra's shortest path in Python using a binary ", "ts": "2026-11-01T01:57:51", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 948.7146249739453, "prompt_tps": 16.925601196101148, "gen_tps": 4.625175756682517, "decode_tps_wall": 4.552927388025366, "itl_ms_p50": 220.21329193376005, "itl_ms_p95": 255.08687505498528, "itl_ms_p99": 265.80637495499104, "itl_ms_mean": 219.6389080638746, "total_s": 14.785965832998045, "peak_mem_gb": 9.161110888, "text": "import heapq\nfrom collections import defaultdict\nimport math\n\ndef dijkstra(graph, start, end):\n # Initialize distances to infinity\n distances = defaultdict(lambda: float('inf'))\n distances[st", "t_start_epoch": 1793516256.346686, "t_end_epoch": 1793516271.136481, "t_first_epoch": 1793516257.2954006, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 8608, "local_pairs": 22112, "layers_remote": 2824, "remote_wait_s": 7.689290434122086, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 52610, "exec_wall_s": 146.54296722088475, "bytes_sent": 281009217, "bytes_recv": 701034948}}}}
141 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "general", "prompt": "Describe the main differences between a Mac Studio and a Mac", "ts": "2026-11-01T01:58:02", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 793.4241249458864, "prompt_tps": 20.263611654446443, "gen_tps": 6.105545976816722, "decode_tps_wall": 6.010172001470645, "itl_ms_p50": 160.93062504660338, "itl_ms_p95": 217.65066601801664, "itl_ms_p99": 235.86608399637043, "itl_ms_mean": 166.38458928551583, "total_s": 11.275653249933384, "peak_mem_gb": 9.161113432, "text": "The Mac Studio and MacBook Pro are both powerful Apple devices designed for high-performance computing, but they differ significantly in terms of architecture, performance, portability, and use cases\u2014", "t_start_epoch": 1793516271.137164, "t_end_epoch": 1793516282.4158342, "t_first_epoch": 1793516271.9305882, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 4159, "local_pairs": 26561, "layers_remote": 1877, "remote_wait_s": 4.8464598739519715, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 54487, "exec_wall_s": 151.5281404076377, "bytes_sent": 291658044, "bytes_recv": 718256292}}}}
142 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "general", "prompt": "Summarize the history of the Rideau Canal in one paragraph.", "ts": "2026-11-01T01:58:17", "prompt_tokens": 14, "gen_tokens": 64, "ttft_ms": 888.1122500170022, "prompt_tps": 15.831335966516948, "gen_tps": 4.535775111168956, "decode_tps_wall": 4.4649165198002505, "itl_ms_p50": 224.13791704457253, "itl_ms_p95": 251.6074589220807, "itl_ms_p99": 269.2011250182986, "itl_ms_mean": 223.9683531742129, "total_s": 14.998118499992415, "peak_mem_gb": 9.161110892, "text": "The Rideau Canal, located in Ontario, Canada, is a historic waterway that stretches approximately 202 kilometers from Ottawa to Lake Ontario. Constructed between 1826 and 1832 under the direction of C", "t_start_epoch": 1793516282.41645, "t_end_epoch": 1793516297.417666, "t_first_epoch": 1793516283.3045623, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 9165, "local_pairs": 20787, "layers_remote": 2866, "remote_wait_s": 7.884127993369475, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 57353, "exec_wall_s": 159.58855458092876, "bytes_sent": 306094028, "bytes_recv": 756080304}}}}
143 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Go function that reads lines from stdin and prints t", "ts": "2026-11-01T01:58:32", "prompt_tokens": 18, "gen_tokens": 64, "ttft_ms": 981.4748329808936, "prompt_tps": 18.387526309682922, "gen_tps": 4.460331601070421, "decode_tps_wall": 4.390647959223199, "itl_ms_p50": 226.1769169708714, "itl_ms_p95": 253.66345897782594, "itl_ms_p99": 264.2245410243049, "itl_ms_mean": 227.75681614358393, "total_s": 15.33015425002668, "peak_mem_gb": 9.161090376, "text": "package main\n\nimport (\n\t\"bufio\"\n\t\"fmt\"\n\t\"os\"\n)\n\nfunc main() {\n\tvar longest string\n\tscanner := bufio.NewScanner(os.Stdin)\n\n\tfor scanner.Scan() {\n\t\tline := scanner.Text()\n\t\tif len(line) > len(longest) {", "t_start_epoch": 1793516297.418133, "t_end_epoch": 1793516312.7519171, "t_first_epoch": 1793516298.399608, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 11180, "local_pairs": 20308, "layers_remote": 3067, "remote_wait_s": 8.318402069504373, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 60420, "exec_wall_s": 168.08417483989615, "bytes_sent": 322165183, "bytes_recv": 802177711}}}}
144 +{"backend": "titan-distributed", "capacity": 64, "placement": "popular:benchmarks/raw/gateF/traces_qwen3_m3u96a", "workers": ["w0"], "rtt_ms": {"w0": [0.7453749421983957, 0.6795410299673676, 0.6044170586392283, 0.5699580069631338, 0.5179588915780187, 0.6379589904099703, 0.4667920293286443, 0.5386249395087361, 0.6382920546457171, 0.6350410403683782]}, "n_remote_experts": 3072, "predictor": "none", "m": 8, "max_tokens": 64, "split": "heldout", "tag": "", "nocache": true, "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_distributed.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --workers w0=192.168.2.85:9412 --placement popular:benchmarks/raw/gateF/traces_qwen3_m3u96a --capacity 64 --nocache --families general,code --split heldout --repeats 2 --max-tokens 64 --out benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "dlogit_vs_mlx_lm": 0.4326338768005371, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short motivational note for someone starting to lear", "ts": "2026-11-01T01:58:46", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 877.9895829502493, "prompt_tps": 18.306623064771383, "gen_tps": 4.9653263117474795, "decode_tps_wall": 4.887763833857396, "itl_ms_p50": 204.08583304379135, "itl_ms_p95": 228.65795902907848, "itl_ms_p99": 273.73579202685505, "itl_ms_mean": 204.592536380958, "total_s": 13.767319374950603, "peak_mem_gb": 9.16111086, "text": "Include a sentence about the beauty of music and the importance of patience.\n\n**Dear Beginner,**\n\nWelcome to your beautiful new journey with the piano. Every great musician once sat where you are now\u2014", "t_start_epoch": 1793516312.7525692, "t_end_epoch": 1793516326.522999, "t_first_epoch": 1793516313.6305587, "cache_hit_rate": 1.0, "cache_misses": 0, "stall_s": 0.0, "remote": {"remote_pairs": 7531, "local_pairs": 23189, "layers_remote": 2805, "remote_wait_s": 7.086085416027345, "failures": 0, "workers": {"w0": {"worker": "w0", "exec_calls": 63225, "exec_wall_s": 175.3545741115231, "bytes_sent": 336726835, "bytes_recv": 833302879}}}}
added benchmarks/raw/tests/distributed_localhost_olmoe_20261101T020003.json +56 −0
@@ -0,0 +1,56 @@
1 +{
2 + "rtt_ms_localhost": [
3 + 0.08691707625985146,
4 + 0.09395892266184092,
5 + 0.08695793803781271,
6 + 0.10974996257573366,
7 + 0.08887494914233685
8 + ],
9 + "prompts": [
10 + {
11 + "prompt": "The capital of Canada is",
12 + "max_abs_dlogit": 0.0,
13 + "argmax_equal": true,
14 + "remote": {
15 + "remote_pairs": 312,
16 + "local_pairs": 328,
17 + "layers_remote": 16,
18 + "remote_wait_s": 0.004074248136021197,
19 + "failures": 0,
20 + "workers": {
21 + "w0": {
22 + "worker": "w0",
23 + "exec_calls": 16,
24 + "exec_wall_s": 0.08466770686209202,
25 + "bytes_sent": 333605,
26 + "bytes_recv": 1282002
27 + }
28 + }
29 + }
30 + },
31 + {
32 + "prompt": "def fibonacci(n):\n \"\"\"Retur",
33 + "max_abs_dlogit": 0.0,
34 + "argmax_equal": true,
35 + "remote": {
36 + "remote_pairs": 1660,
37 + "local_pairs": 1668,
38 + "layers_remote": 56,
39 + "remote_wait_s": 0.012931160512380302,
40 + "failures": 0,
41 + "workers": {
42 + "w0": {
43 + "worker": "w0",
44 + "exec_calls": 56,
45 + "exec_wall_s": 0.16169150278437883,
46 + "bytes_sent": 1723772,
47 + "bytes_recv": 6808122
48 + }
49 + }
50 + }
51 + }
52 + ],
53 + "greedy_equal": true,
54 + "greedy_ref": "Ottawa.\n\nOttawa is the capital of Canada, and it is located in the province of Ontario. It is",
55 + "greedy_titan": "Ottawa.\n\nOttawa is the capital of Canada, and it is located in the province of Ontario. It is"
56 +}
\ No newline at end of file
added benchmarks/raw/tests/fused_upgate_olmoe_2026-11-01T020004.json +26 −0
@@ -0,0 +1,26 @@
1 +{
2 + "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b",
3 + "ts": "2026-11-01T02:00:04",
4 + "cases": [
5 + {
6 + "capacity": 64,
7 + "prompt": "The capital of Canada is",
8 + "max_abs_dlogit": 0.0
9 + },
10 + {
11 + "capacity": 64,
12 + "prompt": "def fibonacci(n):\n \"\"\"Retur",
13 + "max_abs_dlogit": 0.0
14 + },
15 + {
16 + "capacity": 16,
17 + "prompt": "The capital of Canada is",
18 + "max_abs_dlogit": 0.0
19 + },
20 + {
21 + "capacity": 16,
22 + "prompt": "def fibonacci(n):\n \"\"\"Retur",
23 + "max_abs_dlogit": 0.0
24 + }
25 + ]
26 +}
\ No newline at end of file
added benchmarks/raw/tests/greedy_olmoe_cap16_20261101T020010.json +14 −0
@@ -0,0 +1,14 @@
1 +[
2 + {
3 + "prompt": "The capital of Canada is",
4 + "ref": "Ottawa.\n\nOttawa is the capital of Canada, and it is located in the province of Ontario. It is situated on the south bank of the Ottawa",
5 + "titan": "Ottawa.\n\nOttawa is the capital of Canada, and it is located in the province of Ontario. It is situated on the south bank of the Ottawa",
6 + "equal": true
7 + },
8 + {
9 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
10 + "ref": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n",
11 + "titan": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n",
12 + "equal": true
13 + }
14 +]
\ No newline at end of file
added benchmarks/raw/tests/numerical_olmoe_cap16_2026-11-01T020006.json +50 −0
@@ -0,0 +1,50 @@
1 +{
2 + "capacity": 16,
3 + "prompts": [
4 + {
5 + "prompt": "The capital of Canada is",
6 + "tokens": 5,
7 + "max_abs_dlogit": 0.0,
8 + "argmax_equal": true,
9 + "hit_rate": 0.2580101180438449,
10 + "misses": 440,
11 + "stall_s": 0.0558101215865463,
12 + "layers_routed": 46
13 + },
14 + {
15 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
16 + "tokens": 21,
17 + "max_abs_dlogit": 0.0,
18 + "argmax_equal": true,
19 + "hit_rate": 0.42512908777969016,
20 + "misses": 1002,
21 + "stall_s": 0.12626520439516753,
22 + "layers_routed": 127
23 + },
24 + {
25 + "prompt": "Q: If a train travels 60 km in 45 minute",
26 + "tokens": 25,
27 + "max_abs_dlogit": 0.0,
28 + "argmax_equal": true,
29 + "hit_rate": 0.41904344010530936,
30 + "misses": 1324,
31 + "stall_s": 0.16621667833533138,
32 + "layers_routed": 161
33 + },
34 + {
35 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent",
36 + "tokens": 26,
37 + "max_abs_dlogit": 0.0,
38 + "argmax_equal": true,
39 + "hit_rate": 0.45278969957081544,
40 + "misses": 1275,
41 + "stall_s": 0.16034187946934253,
42 + "layers_routed": 164
43 + }
44 + ],
45 + "tol": 0.001,
46 + "ts": "2026-11-01T02:00:06",
47 + "max_abs_dlogit_worst": 0.0,
48 + "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b",
49 + "model_type": "olmoe"
50 +}
\ No newline at end of file
added benchmarks/raw/tests/numerical_olmoe_cap32_2026-11-01T020006.json +50 −0
@@ -0,0 +1,50 @@
1 +{
2 + "capacity": 32,
3 + "prompts": [
4 + {
5 + "prompt": "The capital of Canada is",
6 + "tokens": 5,
7 + "max_abs_dlogit": 0.0,
8 + "argmax_equal": true,
9 + "hit_rate": 0.0,
10 + "misses": 430,
11 + "stall_s": 0.05420312483329326,
12 + "layers_routed": 16
13 + },
14 + {
15 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
16 + "tokens": 21,
17 + "max_abs_dlogit": 0.0,
18 + "argmax_equal": true,
19 + "hit_rate": 0.6060606060606061,
20 + "misses": 429,
21 + "stall_s": 0.054924083640798926,
22 + "layers_routed": 40
23 + },
24 + {
25 + "prompt": "Q: If a train travels 60 km in 45 minute",
26 + "tokens": 25,
27 + "max_abs_dlogit": 0.0,
28 + "argmax_equal": true,
29 + "hit_rate": 0.6090458488228006,
30 + "misses": 631,
31 + "stall_s": 0.08155062305741012,
32 + "layers_routed": 58
33 + },
34 + {
35 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent",
36 + "tokens": 26,
37 + "max_abs_dlogit": 0.0,
38 + "argmax_equal": true,
39 + "hit_rate": 0.6224188790560472,
40 + "misses": 512,
41 + "stall_s": 0.07084849884267896,
42 + "layers_routed": 49
43 + }
44 + ],
45 + "tol": 0.001,
46 + "ts": "2026-11-01T02:00:06",
47 + "max_abs_dlogit_worst": 0.0,
48 + "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b",
49 + "model_type": "olmoe"
50 +}
\ No newline at end of file
added benchmarks/raw/tests/numerical_olmoe_cap64_2026-11-01T020005.json +50 −0
@@ -0,0 +1,50 @@
1 +{
2 + "capacity": 64,
3 + "prompts": [
4 + {
5 + "prompt": "The capital of Canada is",
6 + "tokens": 5,
7 + "max_abs_dlogit": 0.0,
8 + "argmax_equal": true,
9 + "hit_rate": 0.0,
10 + "misses": 430,
11 + "stall_s": 0.05429058114532381,
12 + "layers_routed": 16
13 + },
14 + {
15 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
16 + "tokens": 21,
17 + "max_abs_dlogit": 0.0,
18 + "argmax_equal": true,
19 + "hit_rate": 0.4637904468412943,
20 + "misses": 348,
21 + "stall_s": 0.043974502477794886,
22 + "layers_routed": 16
23 + },
24 + {
25 + "prompt": "Q: If a train travels 60 km in 45 minute",
26 + "tokens": 25,
27 + "max_abs_dlogit": 0.0,
28 + "argmax_equal": true,
29 + "hit_rate": 0.823076923076923,
30 + "misses": 138,
31 + "stall_s": 0.01845787698403001,
32 + "layers_routed": 16
33 + },
34 + {
35 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent",
36 + "tokens": 26,
37 + "max_abs_dlogit": 0.0,
38 + "argmax_equal": true,
39 + "hit_rate": 0.9234828496042217,
40 + "misses": 58,
41 + "stall_s": 0.008064083172939718,
42 + "layers_routed": 16
43 + }
44 + ],
45 + "tol": 0.001,
46 + "ts": "2026-11-01T02:00:05",
47 + "max_abs_dlogit_worst": 0.0,
48 + "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b",
49 + "model_type": "olmoe"
50 +}
\ No newline at end of file
added benchmarks/raw/tests/numerical_olmoe_cap8_2026-11-01T020007.json +50 −0
@@ -0,0 +1,50 @@
1 +{
2 + "capacity": 8,
3 + "prompts": [
4 + {
5 + "prompt": "The capital of Canada is",
6 + "tokens": 5,
7 + "max_abs_dlogit": 0.0,
8 + "argmax_equal": true,
9 + "hit_rate": 0.1625,
10 + "misses": 536,
11 + "stall_s": 0.06750592112075537,
12 + "layers_routed": 80
13 + },
14 + {
15 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
16 + "tokens": 21,
17 + "max_abs_dlogit": 0.0,
18 + "argmax_equal": true,
19 + "hit_rate": 0.4425981873111782,
20 + "misses": 1476,
21 + "stall_s": 0.1891609689919278,
22 + "layers_routed": 331
23 + },
24 + {
25 + "prompt": "Q: If a train travels 60 km in 45 minute",
26 + "tokens": 25,
27 + "max_abs_dlogit": 0.0,
28 + "argmax_equal": true,
29 + "hit_rate": 0.375,
30 + "misses": 1995,
31 + "stall_s": 0.2539524263702333,
32 + "layers_routed": 399
33 + },
34 + {
35 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent",
36 + "tokens": 26,
37 + "max_abs_dlogit": 0.0,
38 + "argmax_equal": true,
39 + "hit_rate": 0.3878048780487805,
40 + "misses": 2008,
41 + "stall_s": 0.25746153946965933,
42 + "layers_routed": 410
43 + }
44 + ],
45 + "tol": 0.001,
46 + "ts": "2026-11-01T02:00:07",
47 + "max_abs_dlogit_worst": 0.0,
48 + "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b",
49 + "model_type": "olmoe"
50 +}
\ No newline at end of file
modified research/LOG.md +10 −0
@@ -158,3 +158,13 @@ Chronological, append-only. Times are local (America/Toronto). Every entry names
158 158 the probe beats demand loading by +32 % (the 64-token runs showed a tie) — see audit CIs; the hit rate reaches 0.84 over longer
159 159 sequences while the per-token install overhead is unchanged. Energy favours prefetch per token in gross terms (the machine is on
160 160 anyway) but not net of idle; all-resident decoding is 4–5× cheaper per token than any cold-tier configuration.
161 +- **Gate F, first sweep discarded for exactness** (`benchmarks/raw/gateF/two_node_qwen3_m3u96a_m3u96b_v1_fp16wire.jsonl`,
162 + `run_gateF_v1_fp16wire.log`): with Qwen3-30B the remote path gave max |Δlogit| = 0.65 (odd) / 0.43 (popular) vs mlx-lm because the
163 + transport cast bfloat16 activations and results to float16 on the wire (OLMoE is float16-native, hence its Δ = 0.0). Fix: activations
164 + and results travel in the model's dtype (bfloat16 as uint16 bit patterns, dtype tag in the header); localhost Qwen3 test on M3U96a now
165 + Δ = 0.0, greedy identical (`benchmarks/raw/tests/m3u96a/pytest_distributed_localhost_qwen3_*`). Throughputs of the discarded sweep,
166 + kept for the record (cold local tier on M3U96a, 64 tokens): none 8.7–8.9 tok/s at 32 slots (stall 65 ms/token), 15–20 at 64 slots;
167 + odd 5.4–5.8 / 4.1; popular 6.6–7.0 / 4.5–5.5; remote wait 70–146 ms/token (48 layers × 1.5–3 ms). Sweep relaunched with the fix.
168 +- Earlier the relaunched sweep had also crashed on "There is no Stream(gpu, N) in current thread": MLX arrays created in reader threads
169 + cannot be evaluated in another thread; fixed by returning numpy from reader threads and creating MLX arrays on the consuming thread
170 + (`titan/store.py` `to_mx`, `titan/prefetch.py`).
modified titan/distributed.py +4 −3
@@ -42,9 +42,10 @@ class RemoteExecutor:
42 42 outs = [None] * len(flat)
43 43 # 1. dispatch every remote batch first (activations as float16 rows)
44 44 if remote:
45 − x_np = np.array(x_flat.astype(mx.float16))
45 + dt = x_flat.dtype
46 + x_np = np.array(x_flat.view(mx.uint16)) if dt == mx.bfloat16 else np.array(x_flat)
46 47 for w, pairs in remote.items():
47 − self.workers[w].exec_async(block.table.layer, x_np, [(r, e) for _, r, e in pairs])
48 + self.workers[w].exec_async(block.table.layer, x_np, [(r, e) for _, r, e in pairs], dtype=str(dt).split(".")[-1])
48 49 self.layers_remote += 1
49 50 # 2. local experts through the slot table (same code path as single node)
50 51 if local:
@@ -81,7 +82,7 @@ class RemoteExecutor:
81 82 for i, (pos, _, _) in enumerate(pairs):
82 83 outs[pos] = yf[i]
83 84 continue
84 − ym = mx.array(y)
85 + ym = mx.array(y).view(mx.bfloat16) if x_flat.dtype == mx.bfloat16 else mx.array(y)
85 86 for i, (pos, _, _) in enumerate(pairs):
86 87 outs[pos] = ym[i]
87 88 self.remote_pairs += len(pairs)
modified transport/client.py +4 −4
@@ -31,9 +31,9 @@ class WorkerClient:
31 31 raise RuntimeError(h)
32 32 return h
33 33
34 − def exec_async(self, layer: int, rows: np.ndarray, pairs):
35 − """Send an exec request (non-blocking on the reply)."""
36 − self.f.send("exec", {"layer": int(layer), "rows": int(rows.shape[0]), "dim": int(rows.shape[1]), "pairs": [[int(r), int(e)] for r, e in pairs]}, rows.tobytes())
34 + def exec_async(self, layer: int, rows: np.ndarray, pairs, dtype: str = "float16"):
35 + """Send an exec request (non-blocking on the reply). rows: float16, or uint16 bit patterns when dtype == 'bfloat16'."""
36 + self.f.send("exec", {"layer": int(layer), "rows": int(rows.shape[0]), "dim": int(rows.shape[1]), "dtype": dtype, "pairs": [[int(r), int(e)] for r, e in pairs]}, rows.tobytes())
37 37 self._t0 = time.perf_counter(); self.rows_sent += rows.shape[0]
38 38
39 39 def exec_wait(self):
@@ -41,7 +41,7 @@ class WorkerClient:
41 41 if kind != "exec_reply":
42 42 raise RuntimeError(h)
43 43 self.exec_calls += 1; self.exec_wall += time.perf_counter() - self._t0
44 − return np.frombuffer(payload, dtype=np.float16).reshape(h["n"], -1), h
44 + return np.frombuffer(payload, dtype=np.uint16 if h.get("dtype") == "bfloat16" else np.float16).reshape(h["n"], -1), h
45 45
46 46 def stats(self):
47 47 return {"worker": self.name, "exec_calls": self.exec_calls, "exec_wall_s": self.exec_wall, "bytes_sent": self.f.bytes_sent, "bytes_recv": self.f.bytes_recv}
modified transport/worker.py +9 −8
@@ -40,14 +40,15 @@ class Worker:
40 40 mx.eval(*[b for t in self.tables.values() for b in t.buf.values()])
41 41 return n
42 42
43 − def execute(self, layer, rows_f16: np.ndarray, pairs):
44 − """rows_f16: [R, D] activations; pairs: list of (row, expert). Returns [len(pairs), D] float16 outputs."""
43 + def execute(self, layer, rows: np.ndarray, pairs, dtype: str = "float16"):
44 + """rows: [R, D] activations (float16, or uint16 bit patterns of bfloat16); pairs: list of (row, expert).
45 + Returns [len(pairs), D] outputs in the same dtype (bfloat16 returned as uint16 bit patterns)."""
45 46 t0 = time.perf_counter()
46 47 tab = self.table(layer)
47 48 experts = [e for _, e in pairs]
48 49 slots = tab.ensure(list(dict.fromkeys(experts)))
49 50 slot_of = dict(zip(dict.fromkeys(experts), slots))
50 − x = mx.array(rows_f16) # [R, D]
51 + x = mx.array(rows).view(mx.bfloat16) if dtype == "bfloat16" else mx.array(rows) # [R, D]
51 52 ridx = mx.array([r for r, _ in pairs], dtype=mx.uint32)
52 53 sidx = mx.array([slot_of[e] for e in experts], dtype=mx.uint32)
53 54 xs = x[ridx] # [P, D]
@@ -58,7 +59,7 @@ class Worker:
58 59 transpose=True, group_size=self.gs, bits=self.bits)
59 60 up = qmm(xe, "up"); gate = qmm(xe, "gate")
60 61 y = qmm(swiglu(gate, up), "down").squeeze(-2).squeeze(-2) # [P, D]
61 − out = np.array(y.astype(mx.float16))
62 + out = np.array(y.view(mx.uint16)) if dtype == "bfloat16" else np.array(y.astype(mx.float16))
62 63 self.exec_count += 1; self.exec_time += time.perf_counter() - t0
63 64 return out
64 65
@@ -97,10 +98,10 @@ def handle(conn, addr, token, w: Worker):
97 98 n = w.load(h["assignments"])
98 99 f.send("load_reply", {"loaded": n, "resident": {str(l): t.resident() for l, t in w.tables.items()}})
99 100 elif kind == "exec":
100 − R, D = h["rows"], h["dim"]
101 − x = np.frombuffer(payload, dtype=np.float16).reshape(R, D)
102 − out = w.execute(h["layer"], x, [tuple(p) for p in h["pairs"]])
103 − f.send("exec_reply", {"layer": h["layer"], "n": out.shape[0], "exec_ms": w.exec_time * 1e3}, out.tobytes())
101 + R, D = h["rows"], h["dim"]; dt = h.get("dtype", "float16")
102 + x = np.frombuffer(payload, dtype=np.uint16 if dt == "bfloat16" else np.float16).reshape(R, D)
103 + out = w.execute(h["layer"], x, [tuple(p) for p in h["pairs"]], dt)
104 + f.send("exec_reply", {"layer": h["layer"], "n": out.shape[0], "dtype": dt, "exec_ms": w.exec_time * 1e3}, out.tobytes())
104 105 else:
105 106 f.send("error", {"msg": f"unexpected {kind}"})
106 107 except (ConnectionError, socket.timeout, OSError) as e:
107 108
Menu

Theme
↑↓ navigate ↵ open esc close / or ⌘K anywhere