Python 88.3%
TypeScript 7.6%
Shell 4.1%
1# FAQ — answering the owner's questions from the atlas data (4 providers)23**Status:** every table on this page is **computed** from `generated/*.json` on 2026-09-18 (the jq/python used is shown so the answer can be re-derived after the next build); generator `scripts/generators/synth/faq.py`. Statuses are the record statuses. Where a question needs interpretation, the interpretation is stated. Providers: `openai`, `anthropic`, `xai`, `gemini`.4**Sources:** `generated/models.json`, `generated/tools.json`, `generated/parameters.json`, `generated/endpoints.json`, `generated/streaming-events.json`, `generated/webhook-events.json`, `generated/pricing.json`, `generated/fragments/headers/anthropic-beta-headers.json`, `generated/examples-manifest.json`, plus the docs pages linked per answer.5**Last verified:** 2026-09-1867Questions: [1](#q1) tool X + structured outputs + caching (Anthropic) · [2](#q2) image + reasoning + web search + streaming (OpenAI) · [2b](#q2b) the same combination on all four providers · [3](#q3) exact JSON for feature Y · [4](#q4) SSE / WebSocket events per call type · [5](#q5) endpoint that creates an agent session · [6](#q6) features needing a beta header / beta path · [7](#q7) which model supports combination Z · [8](#q8) which provider offers X and at what price · [9](#q9) where things live in the atlas89## Q1. Which Anthropic model accepts tool X **with** structured outputs **and** prompt caching? <a id="q1"></a>1011Interpretation: model records where `capabilities.structured_outputs_json`, `capabilities.strict_tool_use` and `capabilities.prompt_caching` are all `true`, crossed with the versioned tool `type`s listed in each model's `tools[]`. Retired ids excluded. All 13 active Claude models satisfy the SO + caching precondition, so the answer is really the tool column.1213```bash14jq -r '.[] | select(.provider=="anthropic" and (.status|index("RETIRED")|not) and .capabilities.structured_outputs_json==true and .capabilities.prompt_caching==true) | [.id, ([.tools[].type]|join(","))] | @tsv' generated/models.json15```1617| Model | Status | SO | strict | cache min tok | 1h TTL | web_search | web_fetch | code_exec | tool_search | mcp_toolset | computer toolset GA | computer beta | browser | bash | text_editor | memory | advisor |18|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|19| `claude-fable-5-1` | `DOCUMENTED` · `LIVE_VERIFIED` | ✔ | ✔ | 512 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ |20| `claude-mythos-5-1` | `DOCUMENTED` · `ACCOUNT_RESTRICTED` · `PREVIEW` | ✔ | ✔ | 512 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ |21| `claude-fable-5` | `DOCUMENTED` · `LIVE_VERIFIED` · `LEGACY` | ✔ | ✔ | 512 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ |22| `claude-mythos-5` | `DOCUMENTED` · `ACCOUNT_RESTRICTED` · `PREVIEW` | ✔ | ✔ | 512 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ |23| `claude-mythos-preview` | `DOCUMENTED` · `ACCOUNT_RESTRICTED` · `DEPRECATED` · `PREVIEW` | ✔ | ✔ | 2048 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✘ | ✘ | ✘ | ✔ | ✔ | ✔ | ✘ |24| `claude-opus-5` | `DOCUMENTED` · `LIVE_VERIFIED` | ✔ | ✔ | 512 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ |25| `claude-opus-4-8` | `DOCUMENTED` · `LIVE_VERIFIED` · `LEGACY` | ✔ | ✔ | 1024 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ |26| `claude-opus-4-7` | `DOCUMENTED` · `LIVE_VERIFIED` · `LEGACY` | ✔ | ✔ | 2048 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✘ | ✔ | ✘ | ✔ | ✔ | ✔ | ✔ |27| `claude-opus-4-6` | `DOCUMENTED` · `LIVE_VERIFIED` · `LEGACY` | ✔ | ✔ | 4096 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✘ | ✔ | ✘ | ✔ | ✔ | ✔ | ✔ |28| `claude-opus-4-5-20251101` | `DOCUMENTED` · `LIVE_VERIFIED` · `LEGACY` | ✔ | ✔ | 4096 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✘ | ✔ | ✘ | ✔ | ✔ | ✔ | ✔ |29| `claude-sonnet-5` | `DOCUMENTED` · `LIVE_VERIFIED` | ✔ | ✔ | 1024 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ |30| `claude-sonnet-4-6` | `DOCUMENTED` · `LIVE_VERIFIED` · `LEGACY` | ✔ | ✔ | 1024 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✘ | ✔ | ✘ | ✔ | ✔ | ✔ | ✔ |31| `claude-sonnet-4-5-20250929` | `DOCUMENTED` · `LIVE_VERIFIED` · `LEGACY` | ✔ | ✔ | 1024 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✘ | ✔ | ✘ | ✔ | ✔ | ✔ | ✔ |32| `claude-haiku-4-5-20251001` | `DOCUMENTED` · `LIVE_VERIFIED` | ✔ | ✔ | 4096 | ✔ | ✔ | ✔ | ✔ | ✔ | ✔ | ✘ | ✔ | ✘ | ✔ | ✔ | ✔ | ✔ |3334Caveats from the records: `strict` is rejected on toolsets (`mcp_toolset`, `computer_toolset_20260801`, `browser_toolset_20260801`) and on programmatic callers; changing `output_config.format` invalidates the prompt cache (docs/anthropic/prompt-caching.md); `tool_choice` `any`/`tool` is 400 on Fable 5.1 / Mythos 5.1; `mcp_toolset` and `advisor_20260301` still need beta headers (`mcp-client-2025-11-20`, `advisor-tool-2026-03-01`); Mythos ids are `ACCOUNT_RESTRICTED` (invite only). Tool-level compatibility lists: `generated/compatibility/anthropic-tool-model-matrix.json`, `generated/compatibility/model-tool-matrix.json`.3536## Q2. Which OpenAI models accept image input + reasoning + web search + streaming? <a id="q2"></a>3738Interpretation: `capabilities.image_in`, `capabilities.reasoning`, `capabilities.tool_web_search` and `capabilities.streaming` all `true` in `generated/models.json`. Web search here means the hosted `web_search` tool on `/v1/responses` (Chat Completions only has search models). Retired ids are listed separately.3940```bash41jq -r '.[] | select(.provider=="openai" and .capabilities.image_in==true and .capabilities.reasoning==true and .capabilities.tool_web_search==true and .capabilities.streaming==true) | [.id, (.status|join(",")), (.context_window|tostring), (.capabilities.reasoning_effort_values//[]|join("/"))] | @tsv' generated/models.json42```4344| Model | Status | Context | Max out | Effort values | Structured outputs | Prompt caching | Code interp. | Computer | MCP | Std price in / out |45|---|---|---|---|---|---|---|---|---|---|---|46| `gpt-5` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | minimal/low/medium/high | ✔ | ✔ | ✔ | ✘ | ✔ | 1.25 / 10.0 |47| `gpt-5-2025-08-07` | `DOCUMENTED` · `LIVE_VERIFIED` · `DEPRECATED` | 400000 | 128000 | minimal/low/medium/high | ✔ | ✔ | ✔ | ✘ | ✔ | 1.25 / 10.0 |48| `gpt-5-mini` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | (model default only) | ✔ | ✘ | ✔ | ✘ | ✔ | 0.25 / 2.0 |49| `gpt-5-mini-2025-08-07` | `DOCUMENTED` · `LIVE_VERIFIED` · `DEPRECATED` | 400000 | 128000 | (model default only) | ✔ | ✘ | ✔ | ✘ | ✔ | 0.25 / 2.0 |50| `gpt-5-nano` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | (model default only) | ✔ | ✔ | ✔ | ✘ | ✔ | 0.05 / 0.4 |51| `gpt-5-nano-2025-08-07` | `DOCUMENTED` · `LIVE_VERIFIED` · `DEPRECATED` | 400000 | 128000 | (model default only) | ✔ | ✔ | ✔ | ✘ | ✔ | 0.05 / 0.4 |52| `gpt-5-pro` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 272000 | high | ✔ | ✘ | ✘ | ✘ | ✔ | 15.0 / 120.0 |53| `gpt-5-pro-2025-10-06` | `DOCUMENTED` · `LIVE_VERIFIED` · `DEPRECATED` | 400000 | 272000 | high | ✔ | ✘ | ✘ | ✘ | ✔ | 15.0 / 120.0 |54| `gpt-5.1` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | none/low/medium/high | ✔ | ✔ | ✔ | ✘ | ✔ | 1.25 / 10.0 |55| `gpt-5.1-2025-11-13` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | none/low/medium/high | ✔ | ✔ | ✔ | ✘ | ✔ | 1.25 / 10.0 |56| `gpt-5.2` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✘ | ✔ | 1.75 / 14.0 |57| `gpt-5.2-2025-12-11` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✘ | ✔ | 1.75 / 14.0 |58| `gpt-5.2-pro` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | (model default only) | ✘ | ✘ | ✘ | ✘ | ✔ | 21.0 / 168.0 |59| `gpt-5.2-pro-2025-12-11` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | (model default only) | ✘ | ✘ | ✘ | ✘ | ✔ | 21.0 / 168.0 |60| `gpt-5.3-codex` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | low/medium/high/xhigh | ✔ | ✔ | ✘ | ✘ | ✘ | 1.75 / 14.0 |61| `gpt-5.4` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✔ | ✔ | 2.5 / 15.0 |62| `gpt-5.4-2026-03-05` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✔ | ✔ | 2.5 / 15.0 |63| `gpt-5.4-mini` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✔ | ✔ | 0.75 / 4.5 |64| `gpt-5.4-mini-2026-03-17` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✔ | ✔ | 0.75 / 4.5 |65| `gpt-5.4-nano` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✘ | ✔ | 0.2 / 1.25 |66| `gpt-5.4-nano-2026-03-17` | `DOCUMENTED` · `LIVE_VERIFIED` | 400000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✘ | ✔ | 0.2 / 1.25 |67| `gpt-5.4-pro` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | medium/high/xhigh | ✘ | ✘ | ✘ | ✔ | ✔ | 30.0 / 180.0 |68| `gpt-5.4-pro-2026-03-05` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | medium/high/xhigh | ✘ | ✘ | ✘ | ✔ | ✔ | 30.0 / 180.0 |69| `gpt-5.5` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✔ | ✔ | 5.0 / 30.0 |70| `gpt-5.5-2026-04-23` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | none/low/medium/high/xhigh | ✔ | ✔ | ✔ | ✔ | ✔ | 5.0 / 30.0 |71| `gpt-5.6` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | none/low/medium/high/xhigh/max | ✔ | ✔ | ✔ | ✔ | ✔ | 4.0 / 20.0 |72| `gpt-5.6-cyber` | `DOCUMENTED` · `ACCOUNT_RESTRICTED` | 400000 | 128000 | (model default only) | ✔ | ✔ | ✔ | ✔ | ✔ | 12.5 / 75.0 |73| `gpt-5.6-luna` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | none/low/medium/high/xhigh/max | ✔ | ✔ | ✔ | ✔ | ✔ | 0.2 / 1.2 |74| `gpt-5.6-sol` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | none/low/medium/high/xhigh/max | ✔ | ✔ | ✔ | ✔ | ✔ | 4.0 / 20.0 |75| `gpt-5.6-terra` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | none/low/medium/high/xhigh/max | ✔ | ✔ | ✔ | ✔ | ✔ | 2.0 / 12.0 |76| `gpt-6-astra` | `DOCUMENTED` · `LIVE_VERIFIED` | 1050000 | 128000 | low/medium/high/xhigh/max | ✔ | ✔ | ✔ | ✔ | ✔ | 10.0 / 50.0 |77| `gpt-daybreak-blue-latest` | `DOCUMENTED` · `ACCOUNT_RESTRICTED` | 1050000 | 128000 | (model default only) | ✔ | ✔ | ✔ | ✔ | ✔ | — / — |78| `gpt-daybreak-red-latest` | `DOCUMENTED` | 400000 | 128000 | (model default only) | ✔ | ✔ | ✔ | ✔ | ✔ | — / — |79| `o3` | `DOCUMENTED` · `LIVE_VERIFIED` | 200000 | 100000 | (model default only) | ✔ | ✔ | ✔ | ✘ | ✔ | 2.0 / 8.0 |80| `o3-2025-04-16` | `DOCUMENTED` · `LIVE_VERIFIED` · `DEPRECATED` | 200000 | 100000 | (model default only) | ✔ | ✔ | ✔ | ✘ | ✔ | 2.0 / 8.0 |81| `o4-mini` | `DOCUMENTED` · `LIVE_VERIFIED` · `DEPRECATED` | 200000 | 100000 | (model default only) | ✔ | ✔ | ✔ | ✘ | ✔ | 1.1 / 4.4 |82| `o4-mini-2025-04-16` | `DOCUMENTED` · `LIVE_VERIFIED` · `DEPRECATED` | 200000 | 100000 | (model default only) | ✔ | ✔ | ✔ | ✘ | ✔ | 1.1 / 4.4 |8384Also matching but **RETIRED** (excluded): `gpt-5-codex`, `gpt-5.1-codex`, `gpt-5.1-codex-max`, `gpt-5.1-codex-mini`, `gpt-5.2-codex`, `o3-deep-research`, `o3-deep-research-2025-06-26`, `o4-mini-deep-research`, `o4-mini-deep-research-2025-06-26`. Caveats: `o4-mini`, `o3-2025-04-16`, `gpt-5-*-2025-08-07` snapshots are `DEPRECATED` with shutdown dates (see docs/openai/deprecations.md); web search is not supported with `gpt-5` `reasoning.effort: minimal` and may degrade with `none` (docs/tools/openai/web-search.md); `gpt-5.6-cyber`, `gpt-daybreak-*` are gated (`ACCOUNT_RESTRICTED`/`DOCUMENTED`).8586## Q2b. The same combination on all four providers — image input + reasoning + vendor web search + structured outputs + streaming <a id="q2b"></a>8788Interpretation per provider (capability keys differ): **openai** `image_in`, `reasoning`, `tool_web_search`, `structured_outputs`, `streaming`; **anthropic** `vision`, (`thinking_adaptive` or `thinking_extended_manual_budget`), `web_search`, `structured_outputs_json` (streaming is universal); **xai** `image_input`, `reasoning`, `web_search`, `structured_outputs`, `streaming`; **gemini** `image_input`, `thinking` (true or a 'Supported…' string), `google_search_grounding`, `structured_output` (streaming is universal). Retired ids, aliases and id-only records excluded; xAI `retired_redirect`/`legacy` kinds and Gemini `alias`/`agent` kinds excluded.8990```bash91jq -r '.[] | select(.provider=="xai" and .kind=="model" and .capabilities.image_input==true and .capabilities.reasoning==true and .capabilities.web_search==true and .capabilities.structured_outputs==true) | .id' generated/models.json92jq -r '.[] | select(.provider=="gemini" and (.kind=="stable" or .kind=="preview") and (.status|index("RETIRED")|not) and .capabilities.image_input==true and (.capabilities.thinking==true or (.capabilities.thinking|type=="string" and startswith("Supported"))) and .capabilities.google_search_grounding==true and .capabilities.structured_output==true) | .id' generated/models.json93```9495| Provider | Matching models (status other than DOCUMENTED/LIVE_*) | Reasoning control | Web search tool | Structured output parameter |96|---|---|---|---|---|97| openai | `gpt-5`, `gpt-5-2025-08-07` (DEPRECATED), `gpt-5-mini`, `gpt-5-mini-2025-08-07` (DEPRECATED), `gpt-5-nano`, `gpt-5-nano-2025-08-07` (DEPRECATED), `gpt-5-pro`, `gpt-5-pro-2025-10-06` (DEPRECATED), `gpt-5.1`, `gpt-5.1-2025-11-13`, `gpt-5.2`, `gpt-5.2-2025-12-11`, `gpt-5.3-codex`, `gpt-5.4`, `gpt-5.4-2026-03-05`, `gpt-5.4-mini`, `gpt-5.4-mini-2026-03-17`, `gpt-5.4-nano`, `gpt-5.4-nano-2026-03-17`, `gpt-5.5`, `gpt-5.5-2026-04-23`, `gpt-5.6`, `gpt-5.6-cyber` (ACCOUNT_RESTRICTED), `gpt-5.6-luna`, `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-6-astra`, `gpt-daybreak-blue-latest` (ACCOUNT_RESTRICTED), `gpt-daybreak-red-latest`, `o3`, `o3-2025-04-16` (DEPRECATED), `o4-mini` (DEPRECATED), `o4-mini-2025-04-16` (DEPRECATED) | `reasoning.effort` (Responses) / `reasoning_effort` (Chat) | `tools[type=web_search]` | `text.format {type: json_schema}` |98| anthropic | `claude-fable-5-1`, `claude-mythos-5-1` (ACCOUNT_RESTRICTED/PREVIEW), `claude-fable-5` (LEGACY), `claude-mythos-5` (ACCOUNT_RESTRICTED/PREVIEW), `claude-mythos-preview` (ACCOUNT_RESTRICTED/DEPRECATED/PREVIEW), `claude-opus-5`, `claude-opus-4-8` (LEGACY), `claude-opus-4-7` (LEGACY), `claude-opus-4-6` (LEGACY), `claude-opus-4-5-20251101` (LEGACY), `claude-sonnet-5`, `claude-sonnet-4-6` (LEGACY), `claude-sonnet-4-5-20250929` (LEGACY), `claude-haiku-4-5-20251001` | `thinking {type: adaptive \| enabled}` + `output_config.effort` | `tools[type=web_search_20260318]` | `output_config.format {type: json_schema}` |99| xai | `grok-4.6` | `reasoning.effort` / `reasoning_effort` (low…xhigh; `none` LIVE_DISCOVERED on grok-4.3; 400 on 4.20-reasoning / build) | `tools[type=web_search]` ($5/1k) + `x_search` | `text.format {type: json_schema}` / `response_format.json_schema` |100| gemini | `gemini-2.5-flash`, `gemini-2.5-pro`, `gemini-2.5-flash-lite` (ACCOUNT_RESTRICTED), `gemini-3-flash-preview` (PREVIEW), `gemini-3.1-pro-preview` (PREVIEW/ACCOUNT_RESTRICTED), `gemini-3.1-pro-preview-customtools` (PREVIEW), `gemini-3.1-flash-lite` (DEPRECATED), `gemini-3.5-flash`, `gemini-3.5-flash-lite`, `gemini-3.6-flash`, `gemini-3.7-flash`, `gemini-3.8-flash`, `gemini-robotics-er-2-preview` (PREVIEW), `gemini-robotics-er-2-streaming-preview` (PREVIEW), `gemini-2.5-flash-preview-09-2025` (PREVIEW/UNVERIFIED) | `generationConfig.thinkingConfig {thinkingLevel \| thinkingBudget, includeThoughts}` | `tools[{googleSearch:{}}]` (ACCOUNT_RESTRICTED on this key; 5,000 free/month on 3.x then $14/1k) | `generationConfig.responseMimeType: application/json` + `responseJsonSchema` / `responseSchema` |101102Reading notes: OpenAI snapshots (`gpt-5.4-2026-03-05`…) appear as separate callable ids; Anthropic aliases (`claude-haiku-4-5`) are excluded (capabilities live on the snapshot); Gemini image-output models (`gemini-3.1-flash-image`, `gemini-3-pro-image`) support Google Search grounding and thinking but not structured output, so they drop out; `gemini-2.5-*` match on paper but are `ACCOUNT_RESTRICTED` ('no longer available to new users'); xAI `grok-4.20-0309-non-reasoning` drops out on `reasoning: false`.103104## Q3. What is the exact JSON to call feature Y? <a id="q3"></a>105106Three lookups, in order of precision:1071081. **Parameter rows** — `generated/parameters.json` (7,862 rows) is keyed by `endpoint` (`METHOD /path`) and dotted `parameter` path (Gemini uses the REST camelCase names, e.g. `generationConfig.thinkingConfig.thinkingLevel`), with `type`, `required`, `default`, `enum`, `beta_header`, `status`, `description`.109110```bash111# every parameter of the Anthropic structured-output block112jq '[.[] | select(.provider=="anthropic" and .endpoint=="POST /v1/messages" and (.parameter|startswith("output_config.format")))]' generated/parameters.json113# the OpenAI equivalent114jq '[.[] | select(.provider=="openai" and .endpoint=="POST /v1/responses" and (.parameter|startswith("text.format")))]' generated/parameters.json115# xAI Responses reasoning + caching knobs116jq '[.[] | select(.provider=="xai" and .endpoint=="POST /v1/responses" and (.parameter|test("^(reasoning|prompt_cache_key|store|previous_response_id|max_turns)")))]' generated/parameters.json117# Gemini thinking and structured-output knobs118jq '[.[] | select(.provider=="gemini" and .endpoint=="POST /v1beta/models/{model}:generateContent" and (.parameter|test("thinkingConfig|responseJsonSchema|responseMimeType|cachedContent")))]' generated/parameters.json119# all tool entry shapes accepted by OpenAI Responses120jq -r '.[] | select(.provider=="openai" and .endpoint=="POST /v1/responses" and (.parameter|test("^tools\\[\\]\\("))) | .parameter' generated/parameters.json121```1221232. **Tool records** — `generated/tools.json` has `parameters_schema`, `result_shape`, `examples{curl,python,typescript}` per exact tool `type` (69 records: OpenAI 19, Anthropic 27, xAI 12, Gemini 11).124125```bash126jq '.[] | select(.type=="web_search_20260318") | {parameters_schema, result_shape, beta_header, examples}' generated/tools.json127jq '.[] | select(.provider=="xai" and .type=="x_search") | {parameters_schema, result_shape, billing}' generated/tools.json128jq '.[] | select(.provider=="gemini" and .type=="googleSearch") | {parameters_schema, result_shape, billing, limitations}' generated/tools.json129```1301313. **Runnable examples** — `generated/examples-manifest.json` maps features to files under `examples/<provider>/<area>/` with their verification status. Selection:132133| Feature | OpenAI | Anthropic | xAI | Gemini |134|---|---|---|---|---|135| Minimal call | `examples/openai/responses/` | `examples/anthropic/messages/` | `examples/xai/responses/minimal.py` (LIVE_VERIFIED) | `examples/gemini/generate-content/minimal.py` (LIVE_VERIFIED) |136| Structured outputs | `examples/openai/responses/structured_output.py` (LIVE_VERIFIED) | `examples/anthropic/structured-output/` | `examples/xai/structured-output/json_schema.sh` (LIVE_VERIFIED) | `examples/gemini/structured-output/json_schema.py` (LIVE_VERIFIED) |137| Streaming | `examples/openai/responses/streaming.py` (LIVE_VERIFIED) | `examples/anthropic/streaming/sdk_stream.py` (LIVE_VERIFIED) | `examples/xai/streaming/responses_stream.py` (LIVE_VERIFIED) | `examples/gemini/streaming/stream.py` (LIVE_VERIFIED) |138| Multi-turn state | `examples/openai/responses/previous_response_id.py` (LIVE_VERIFIED) | `examples/shared/tool-loop/anthropic_tool_loop.py` (LIVE_VERIFIED) | `examples/xai/responses/compact.py` (LIVE_VERIFIED) | `examples/gemini/generate-content/system_multiturn.py` (LIVE_VERIFIED) |139| Tool loop | `examples/shared/tool-loop/openai_tool_loop.py` (not in manifest) | `examples/shared/tool-loop/anthropic_tool_loop.py` (LIVE_VERIFIED) | `examples/xai/responses/tool_loop.py` (LIVE_VERIFIED) | `examples/shared/tool-loop/gemini_tool_loop.py` (LIVE_VERIFIED) |140| Web search | `examples/openai/tools/web-search/` | `examples/anthropic/web-search/web_search.py` (LIVE_VERIFIED) | `examples/xai/tools/web-search/web_search.py` (LIVE_VERIFIED) | `examples/gemini/tools/google-search/grounding.py` (FAILED_VERIFICATION) |141| Code execution | `examples/openai/tools/code-interpreter/` | `examples/anthropic/code-execution/code_execution.py` (LIVE_VERIFIED) | `examples/xai/tools/code-execution/code_interpreter.py` (LIVE_VERIFIED) | `examples/gemini/tools/code-execution/code_execution.py` (LIVE_VERIFIED) |142| MCP | `examples/openai/tools/mcp-and-connectors/` | `examples/anthropic/mcp/mcp_connector.py` (LIVE_VERIFIED) | `examples/xai/tools/mcp/deepwiki.py` (LIVE_VERIFIED) | `docs/tools/gemini/mcp.md` |143| File search / RAG | `docs/openai/vector-stores.md` | `docs/anthropic/citations.md` | `examples/xai/tools/collections-search/file_search.py` (LIVE_VERIFIED) | `examples/gemini/file-search/file_search_lifecycle.py` (LIVE_VERIFIED) |144| Batch | `examples/openai/batch/batch-lifecycle.py` (LIVE_VERIFIED) | `examples/anthropic/batch/batch_lifecycle.py` (LIVE_VERIFIED) | `examples/xai/batches/lifecycle.py` (LIVE_VERIFIED) | `examples/gemini/batch/batch_inline.py` (UNVERIFIED) |145| Agents / sessions | `examples/openai/agents/` | `examples/anthropic/agents/create_agent_session_stream.py` (LIVE_VERIFIED) | `examples/xai/responses/tool_loop.py` (LIVE_VERIFIED) | `examples/gemini/interactions/interactions_basic.py` (LIVE_VERIFIED) |146| Prompt / context caching | `docs/openai/prompt-caching.md §5 (live pair)` | `examples/anthropic/prompt-caching/` | `docs/xai/prompt-caching.md` | `examples/gemini/context-caching/cache_lifecycle.py` (LIVE_VERIFIED) |147| Reasoning / thinking | `examples/openai/responses/reasoning_summary.py` (LIVE_VERIFIED) | `examples/anthropic/thinking/` | `docs/xai/reasoning.md` | `examples/gemini/thinking/thinking.py` (LIVE_VERIFIED) |148| Image input | `examples/openai/responses/image_input.py` (LIVE_VERIFIED) | `examples/anthropic/vision/` | `examples/xai/responses/image_input.sh` (LIVE_VERIFIED) | `examples/gemini/multimodal/inline_image_pdf.py` (LIVE_VERIFIED) |149| Realtime / Live voice | `docs/openai/realtime.md` | — | `docs/xai/voice.md` | `examples/gemini/live/live_ws_audio.py` (LIVE_VERIFIED) |150| Image generation | `docs/openai/images.md` | — | `docs/xai/images.md` | `examples/gemini/image-generation/generate.py` (UNVERIFIED) |151| Embeddings | `docs/openai/embeddings.md` | — | `docs/xai/index.md (ACCOUNT_RESTRICTED)` | `examples/gemini/embeddings/embed.py` (LIVE_VERIFIED) |152| Skills | `docs/openai/skills-api.md` | `examples/anthropic/skills/skills_lifecycle.py` (LIVE_VERIFIED) | `docs/xai/skills-api.md (ACCOUNT_RESTRICTED)` | — |153| Webhook verification | `examples/openai/webhooks/offline_test.py` (LIVE_VERIFIED) | — | `docs/xai/voice.md (SIP `realtime.call.incoming`)` | `docs/gemini/interactions-api.md (`webhook_config`)` |154155Minimal verified bodies (from the live probes recorded in the docs):156157```json158// OpenAI — POST /v1/responses (structured output, LIVE_VERIFIED gpt-5.4-nano)159{"model":"gpt-5.4-nano","input":"Reply with OK.","max_output_tokens":32,160 "text":{"format":{"type":"json_schema","name":"ok_reply","strict":true,161 "schema":{"type":"object","properties":{"answer":{"type":"string"}},"required":["answer"],"additionalProperties":false}}}}162```163```json164// Anthropic — POST /v1/messages (structured output, LIVE_VERIFIED claude-haiku-4-5-20251001)165{"model":"claude-haiku-4-5-20251001","max_tokens":100,166 "messages":[{"role":"user","content":"Reply with OK."}],167 "output_config":{"format":{"type":"json_schema",168 "schema":{"type":"object","properties":{"answer":{"type":"string"}},"required":["answer"],"additionalProperties":false}}}}169```170```json171// xAI — POST /v1/responses (structured output, LIVE_VERIFIED grok-4.3; reasoning tokens are billed on every call)172{"model":"grok-4.3","input":"Reply with OK.","max_output_tokens":64,"reasoning":{"effort":"low"},173 "text":{"format":{"type":"json_schema","name":"ok_reply","strict":true,174 "schema":{"type":"object","properties":{"answer":{"type":"string"}},"required":["answer"],"additionalProperties":false}}}}175```176```json177// Gemini — POST /v1beta/models/gemini-3.5-flash-lite:generateContent (structured output, LIVE_VERIFIED; header x-goog-api-key)178{"contents":[{"role":"user","parts":[{"text":"Reply with OK."}]}],179 "generationConfig":{"maxOutputTokens":64,"thinkingConfig":{"thinkingLevel":"minimal"},180 "responseMimeType":"application/json",181 "responseJsonSchema":{"type":"object","properties":{"answer":{"type":"string"}},"required":["answer"]}}}182```183184Per-topic parameter mappings with full JSON quads: [state-management](comparisons/state-management.md), [tool-execution](comparisons/tool-execution.md), [caching-and-reasoning](comparisons/caching-and-reasoning.md), [streaming](comparisons/streaming.md), [agents-platforms](comparisons/agents-platforms.md), [realtime-and-media](comparisons/realtime-and-media.md).185186## Q4. Which SSE / WebSocket events can arrive during this call? <a id="q4"></a>187188Computed from `generated/streaming-events.json` (506 records: openai 282, anthropic 68, xai 94, gemini 62; `api` field = call type, `direction` = server→client unless noted). Event names are listed exactly as recorded; ✔ marks events observed live in this run (`LIVE_VERIFIED`).189190```bash191jq -r '.[] | select(.provider=="xai" and .api=="responses") | .event' generated/streaming-events.json192jq -r '.[] | select(.provider=="gemini" and .api=="live" and .direction=="server→client") | .event' generated/streaming-events.json193```194195| Provider | Call type (`api`) | Direction | # | Events (✔ = LIVE_VERIFIED) |196|---|---|---|---|---|197| openai | `responses` | server→client | 63 | `response.audio.delta`, `response.audio.done`, `response.audio.transcript.delta`, `response.audio.transcript.done`, `response.code_interpreter_call_code.delta`, `response.code_interpreter_call_code.done`, `response.code_interpreter_call.completed`, `response.code_interpreter_call.in_progress`, `response.code_interpreter_call.interpreting`, `response.compaction.compacting`, ✔`response.completed`, ✔`response.content_part.added`, ✔`response.content_part.done`, ✔`response.created`, `error`, ✔`response.file_search_call.completed`, ✔`response.file_search_call.in_progress`, ✔`response.file_search_call.searching`, ✔`response.function_call_arguments.delta`, ✔`response.function_call_arguments.done`, ✔`response.shell_call_command.added`, ✔`response.shell_call_command.delta`, ✔`response.shell_call_command.done`, `response.shell_call_output_content.delta`, `response.shell_call_output_content.done`, ✔`response.in_progress`, `response.failed`, `response.incomplete`, ✔`response.output_item.added`, ✔`response.output_item.done`, `response.reasoning_summary_part.added`, `response.reasoning_summary_part.done`, `response.reasoning_summary_text.delta`, `response.reasoning_summary_text.done`, `response.reasoning_text.delta`, `response.reasoning_text.done`, `response.refusal.delta`, `response.refusal.done`, ✔`response.output_text.delta`, ✔`response.output_text.done`, ✔`response.web_search_call.completed`, ✔`response.web_search_call.in_progress`, ✔`response.web_search_call.searching`, `response.image_generation_call.completed`, `response.image_generation_call.generating`, `response.image_generation_call.in_progress`, `response.image_generation_call.partial_image`, ✔`response.mcp_call_arguments.delta`, ✔`response.mcp_call_arguments.done`, ✔`response.mcp_call.completed`, `response.mcp_call.failed`, ✔`response.mcp_call.in_progress`, ✔`response.mcp_list_tools.completed`, `response.mcp_list_tools.failed`, ✔`response.mcp_list_tools.in_progress`, `response.output_text.annotation.added`, ✔`response.queued`, ✔`response.custom_tool_call_input.delta`, ✔`response.custom_tool_call_input.done`, ✔`_sequence_observed`, ✔`response.apply_patch_call_operation_diff.delta`, ✔`response.apply_patch_call_operation_diff.done`, ✔`response.output_item.added / response.output_item.done (tool items)` |198| openai | `responses-websocket` | server→client | 69 | `response.audio.delta`, `response.audio.done`, `response.audio.transcript.delta`, `response.audio.transcript.done`, `response.code_interpreter_call_code.delta`, `response.code_interpreter_call_code.done`, `response.code_interpreter_call.completed`, `response.code_interpreter_call.in_progress`, `response.code_interpreter_call.interpreting`, `response.compaction.compacting`, `response.completed`, `response.content_part.added`, `response.content_part.done`, `response.created`, `response.file_search_call.completed`, `response.file_search_call.in_progress`, `response.file_search_call.searching`, `response.function_call_arguments.delta`, `response.function_call_arguments.done`, `response.shell_call_command.added`, `response.shell_call_command.delta`, `response.shell_call_command.done`, `response.shell_call_output_content.delta`, `response.shell_call_output_content.done`, `response.in_progress`, `response.failed`, `response.incomplete`, `response.output_item.added`, `response.output_item.done`, `response.reasoning_summary_part.added`, `response.reasoning_summary_part.done`, `response.reasoning_summary_text.delta`, `response.reasoning_summary_text.done`, `response.reasoning_text.delta`, `response.reasoning_text.done`, `response.refusal.delta`, `response.refusal.done`, `response.output_text.delta`, `response.output_text.done`, `response.web_search_call.completed`, `response.web_search_call.in_progress`, `response.web_search_call.searching`, `response.image_generation_call.completed`, `response.image_generation_call.generating`, `response.image_generation_call.in_progress`, `response.image_generation_call.partial_image`, `response.mcp_call_arguments.delta`, `response.mcp_call_arguments.done`, `response.mcp_call.completed`, `response.mcp_call.failed`, `response.mcp_call.in_progress`, `response.mcp_list_tools.completed`, `response.mcp_list_tools.failed`, `response.mcp_list_tools.in_progress`, `response.output_text.annotation.added`, `response.queued`, `response.custom_tool_call_input.delta`, `response.custom_tool_call_input.done`, `error`, `response.steer.accepted`, `response.steer.pending`, `response.steer.failed`, `response.inject.created`, `response.inject.failed`, `error (code=previous_response_not_found)`, `error (code=invalid_stream_id)`, `error (code=websocket_stream_limit_reached)`, `error (code=websocket_connection_limit_reached)`, `_connection_limits` |199| openai | `responses-websocket` | client→server | 3 | `response.create`, `response.steer`, `response.inject` |200| openai | `chat_completions` | server→client | 2 | ✔`chat.completion.chunk`, ✔`[DONE]` |201| openai | `completions` | server→client | 1 | ✔`text_completion (streamed)` |202| openai | `realtime` | server→client | 45 | ✔`session.created`, ✔`session.updated`, ✔`conversation.item.added`, ✔`conversation.item.done`, `conversation.item.retrieved`, `conversation.item.input_audio_transcription.completed`, `conversation.item.input_audio_transcription.delta`, `conversation.item.input_audio_transcription.segment`, `conversation.item.input_audio_transcription.failed`, `conversation.item.truncated`, `conversation.item.deleted`, `input_audio_buffer.committed`, `input_audio_buffer.dtmf_event_received`, `input_audio_buffer.cleared`, `input_audio_buffer.speech_started`, `input_audio_buffer.speech_stopped`, `input_audio_buffer.timeout_triggered`, `output_audio_buffer.started`, `output_audio_buffer.stopped`, `output_audio_buffer.cleared`, ✔`response.created`, ✔`response.done`, ✔`response.output_item.added`, ✔`response.output_item.done`, ✔`response.content_part.added`, ✔`response.content_part.done`, ✔`response.output_text.delta`, ✔`response.output_text.done`, `response.output_audio_transcript.delta`, `response.output_audio_transcript.done`, `response.output_audio.delta`, `response.output_audio.done`, `response.function_call_arguments.delta`, `response.function_call_arguments.done`, `response.mcp_call_arguments.delta`, `response.mcp_call_arguments.done`, `response.mcp_call.in_progress`, `response.mcp_call.completed`, `response.mcp_call.failed`, `mcp_list_tools.in_progress`, `mcp_list_tools.completed`, `mcp_list_tools.failed`, `rate_limits.updated`, `conversation.created`, `conversation.item.created` |203| openai | `realtime` | client→server | 11 | ✔`session.update`, `input_audio_buffer.append`, `input_audio_buffer.commit`, `input_audio_buffer.clear`, ✔`conversation.item.create`, `conversation.item.retrieve`, `conversation.item.truncate`, `conversation.item.delete`, ✔`response.create`, `response.cancel`, `output_audio_buffer.clear` |204| openai | `realtime-translation` | server→client | 6 | `session.created`, `session.updated`, `session.closed`, `session.input_transcript.delta`, `session.output_transcript.delta`, `session.output_audio.delta` |205| openai | `realtime-translation` | client→server | 3 | `session.update`, `session.input_audio_buffer.append`, `session.close` |206| openai | `live` | server→client | 20 | `session.started`, `session.updated`, `session.input_audio.muted`, `session.input_audio.unmuted`, `session.instructions.appended`, `session.thinking.appended`, `session.commentary.appended`, `session.output_audio.delta`, `session.input_transcript.delta`, `session.output_transcript.delta`, `session.delegation.created`, `response.event`, `session.usage.updated`, `session.closed`, `session.input_audio.append`, `transport.dtmf.received`, `transport.dtmf.send`, `transport.ringing`, `transport.answered`, `transport.failed` |207| openai | `live` | client→server | 11 | `session.start`, `session.update`, `session.input_audio.append`, `session.input_audio.mute`, `session.input_audio.unmute`, `session.instructions.append`, `session.thinking.append`, `session.commentary.append`, `response.item.create`, `response.create`, `session.close` |208| openai | `audio-transcriptions` | server→client | 3 | ✔`transcript.text.segment`, ✔`transcript.text.delta`, ✔`transcript.text.done` |209| openai | `audio-speech` | server→client | 2 | ✔`speech.audio.delta`, ✔`speech.audio.done` |210| anthropic | `messages` | server→client | 23 | ✔`content_block_start (thinking)`, ✔`content_block_delta: thinking_delta`, ✔`content_block_delta: signature_delta`, `content_block_start (redacted_thinking)`, ✔`content_block_delta: citations_delta`, `content_block_start/delta/stop (compaction, threshold)`, `content_block_start (compaction, on-demand)`, ✔`message_delta (usage.output_tokens_details / context_management)`, ✔`message_start (usage cache fields / input_transformations)`, ✔`content_block_delta: text_delta (structured outputs)`, ✔`message_start`, ✔`content_block_start`, ✔`content_block_delta / text_delta`, ✔`content_block_delta / input_json_delta`, ✔`content_block_delta / thinking_delta`, ✔`content_block_delta / signature_delta`, `content_block_delta / citations_delta`, ✔`content_block_stop`, ✔`message_delta`, ✔`message_stop`, ✔`ping`, `error`, `(unknown event types)` |211| anthropic | `managed-agents` | server→client | 37 | `user.message`, `user.interrupt`, `user.tool_confirmation`, `user.custom_tool_result`, `agent.custom_tool_use`, `agent.message`, `agent.thinking`, `agent.mcp_tool_use`, `agent.mcp_tool_result`, `agent.tool_use`, `agent.tool_result`, `agent.thread_message_received`, `agent.thread_message_sent`, `agent.thread_context_compacted`, `session.error`, `session.status_rescheduled`, `session.status_running`, `session.status_idle`, `session.status_terminated`, `session.thread_created`, `span.outcome_evaluation_start`, `span.outcome_evaluation_end`, `span.model_request_start`, `span.model_request_end`, `span.outcome_evaluation_ongoing`, `user.define_outcome`, `session.deleted`, `session.thread_status_running`, `session.thread_status_idle`, `session.thread_status_terminated`, `user.tool_result`, `session.thread_status_rescheduled`, `session.updated`, `event_start`, `event_delta`, `system.message`, `session.usage` |212| anthropic | `managed-agents` | client→server | 7 | `user.message`, `user.interrupt`, `user.tool_confirmation`, `user.custom_tool_result`, `user.define_outcome`, `user.tool_result`, `system.message` |213| xai | `responses` | server→client | 24 | ✔`response.created`, ✔`response.in_progress`, ✔`response.output_item.added`, ✔`response.reasoning_summary_part.added`, ✔`response.reasoning_summary_text.delta`, ✔`response.reasoning_summary_text.done`, ✔`response.reasoning_summary_part.done`, ✔`response.output_item.done`, ✔`response.content_part.added`, ✔`response.output_text.delta`, ✔`response.output_text.done`, ✔`response.content_part.done`, ✔`response.function_call_arguments.delta`, ✔`response.function_call_arguments.done`, ✔`response.code_interpreter_call.in_progress`, ✔`response.code_interpreter_call_code.delta`, ✔`response.code_interpreter_call_code.done`, ✔`response.code_interpreter_call.interpreting`, ✔`response.code_interpreter_call.completed`, ✔`response.completed`, `response.reasoning_text.delta`, `response.web_search_call.in_progress / .searching / .completed`, `response.mcp_call.* / response.file_search_call.*`, `response.incomplete / response.failed / error` |214| xai | `chat_completions` | server→client | 3 | ✔`chat.completion.chunk`, ✔`chat.completion.chunk (usage chunk)`, ✔`[DONE]` |215| xai | `messages` | server→client | 6 | ✔`message_start`, ✔`content_block_start`, ✔`content_block_delta`, ✔`content_block_stop`, ✔`message_delta`, ✔`message_stop` |216| xai | `realtime` | server→client | 39 | ✔`session.created`, ✔`conversation.created`, ✔`session.updated`, `input_audio_buffer.speech_started`, `input_audio_buffer.speech_stopped`, `input_audio_buffer.committed`, `input_audio_buffer.timeout_triggered`, `input_audio_buffer.cleared`, `conversation.item.deleted`, ✔`conversation.item.added`, `conversation.item.truncated`, `conversation.item.input_audio_transcription.completed`, `conversation.item.input_audio_transcription.updated`, `input_audio_buffer.dtmf_event_received`, ✔`response.created`, ✔`response.output_item.added`, ✔`response.output_item.done`, ✔`response.content_part.added`, ✔`response.content_part.done`, ✔`response.output_audio_transcript.delta`, ✔`response.output_audio_transcript.done`, ✔`response.output_audio.delta`, ✔`response.output_audio.done`, `response.text.delta`, `response.output_text.delta`, `response.function_call_arguments.delta`, `response.function_call_arguments.done`, `mcp_list_tools.in_progress`, `mcp_list_tools.completed`, `mcp_list_tools.failed`, `response.mcp_call_arguments.delta`, `response.mcp_call_arguments.done`, `response.mcp_call.in_progress`, `response.mcp_call.completed`, `response.mcp_call.failed`, ✔`response.done`, `error`, `ping`, `response.audio.delta` |217| xai | `realtime` | client→server | 9 | ✔`session.update`, `input_audio_buffer.append`, `input_audio_buffer.commit`, ✔`conversation.item.create`, `input_audio_buffer.clear`, `conversation.item.delete`, `conversation.item.truncate`, ✔`response.create`, `response.cancel` |218| xai | `realtime` | server→client (webhook) | 1 | `realtime.call.incoming` |219| xai | `tts` | client→server | 2 | `text.delta`, `text.done` |220| xai | `tts` | server→client | 3 | `audio.delta`, `audio.done`, `error` |221| xai | `stt` | client→server | 3 | `Binary frame (audio)`, `finalize`, `audio.done` |222| xai | `stt` | server→client | 4 | `transcript.created`, `transcript.partial`, `transcript.done`, `error` |223| gemini | `generate-content` | server→client | 8 | ✔`sse_framing`, ✔`json_array_framing`, ✔`chunk`, ✔`thought_chunk`, ✔`final_chunk`, ✔`structured_output_chunks`, `prompt_blocked`, ✔`single_event_alt_sse_on_generateContent` |224| gemini | `interactions` | server→client | 15 | ✔`interaction.created`, ✔`interaction.status_update`, `interaction.in_progress`, `interaction.requires_action`, ✔`step.start`, `step.delta`, ✔`step.stop`, ✔`interaction.completed`, `error`, ✔`done`, `interaction.start`, `content.start`, `content.delta`, `content.stop`, `interaction.complete` |225| gemini | `live` | server→client | 21 | ✔`setupComplete`, ✔`serverContent`, ✔`serverContent.modelTurn`, ✔`serverContent.generationComplete`, ✔`serverContent.turnComplete`, `serverContent.interactionStatus`, `serverContent.interrupted`, `serverContent.groundingMetadata`, `serverContent.inputTranscription`, `serverContent.interimInputTranscription`, ✔`serverContent.outputTranscription`, `serverContent.urlContextMetadata`, `serverContent.waitingForInput`, `serverContent.speechState`, ✔`toolCall`, `toolCallCancellation`, ✔`usageMetadata`, `goAway`, ✔`sessionResumptionUpdate`, `voiceActivity`, `voiceActivityDetectionSignal` |226| gemini | `live` | client→server | 11 | ✔`setup`, ✔`clientContent`, `realtimeInput`, `realtimeInput.audio`, `realtimeInput.video`, `realtimeInput.text`, `realtimeInput.activityStart`, `realtimeInput.activityEnd`, `realtimeInput.audioStreamEnd`, `realtimeInput.mediaChunks`, `toolResponse` |227| gemini | `lyria-realtime` | server→client | 3 | `setupComplete`, `serverContent.audioChunks[]`, `filteredPrompt` |228| gemini | `lyria-realtime` | client→server | 4 | `setup`, `clientContent`, `musicGenerationConfig`, `playbackControl` |229| anthropic | `complete-legacy` | server→client | 1 | `completion (legacy /v1/complete)` |230| openai | `agents` | server→client | 31 | `error`, `agent.session.environment.ready`, `agent.session.environment.reset`, `agent.output.command_execution_output.delta`, ✔`agent.session.created`, ✔`agent.session.turn.created`, ✔`agent.session.turn.in_progress`, ✔`agent.session.turn.completed`, `agent.session.turn.failed`, `agent.session.turn.cancelled`, ✔`agent.session.turn.item.added`, ✔`agent.session.idle`, ✔`agent.session.in_progress`, `agent.session.requires_action`, `agent.session.failed`, `agent.session.environment.pending`, `agent.session.environment.connected`, `agent.session.environment.disconnected`, `agent.session.environment.failed`, `agent.session.subagent.created`, `agent.session.subagent.active`, `agent.session.subagent.closed`, ✔`agent.session.turn.item.done`, ✔`agent.session.turn.content_part.added`, ✔`agent.session.turn.content_part.done`, ✔`agent.session.turn.output_text.delta`, ✔`agent.session.turn.output_text.done`, `agent.session.turn.reasoning_summary_part.added`, `agent.session.turn.reasoning_summary_part.done`, `agent.session.turn.reasoning_summary_text.delta`, `agent.session.turn.reasoning_summary_text.done` |231| openai | `agents` | client→server | 3 | ✔`agent.session.input.message`, ✔`agent.session.input.cancel`, `agent.session.input.tool_result` |232| openai | `agents-webhooks` | server→client (webhook) | 5 | `agent.session.created`, `agent.session.action_required`, `agent.session.in_progress`, `agent.session.idle`, `agent.session.failed` |233| openai | `images` | server→client | 4 | `image_generation.partial_image`, `image_generation.completed`, `image_edit.partial_image`, `image_edit.completed` |234235How to read the four core streams: **OpenAI Responses** is `event: <type>` + `data:` with `sequence_number`, terminal `response.completed|failed|incomplete|error`, no `[DONE]`; **Anthropic Messages** is `event: <type>` + `data:` (`data.type == event`), fixed skeleton `message_start → content_block_start/delta/stop × N → message_delta → message_stop`, `ping` anywhere, no `[DONE]`; **xAI** reuses both shapes — `/v1/responses` emits the OpenAI Responses event names (with `sequence_number`, no `[DONE]`), `/v1/chat/completions` emits data-only `chat.completion.chunk` + `data: [DONE]`, and `/v1/messages` emits the Anthropic skeleton (no `ping`); **Gemini** `streamGenerateContent` has **no event names at all** — a JSON array of `GenerateContentResponse` chunks by default or `data:` lines with `?alt=sse` (thought parts flagged `thought: true`, end detected by `finishReason`), while the Interactions API uses named events (`interaction.start … content.delta … interaction.complete`). Details: [streaming comparison](comparisons/streaming.md).236237The Anthropic `messages` group is the *advanced* fragment (thinking/citations/compaction variants of the same core events); the `POST /v1/messages (stream=true)` group is the core catalogue — both describe one stream. Voice WebSockets (OpenAI Realtime/Live, xAI realtime/tts/stt, Gemini Live/Lyria RealTime) are bidirectional message families, not SSE.238239## Q5. Which endpoint creates an agent session on each provider? <a id="q5"></a>240241```bash242jq '.[] | select((.provider=="openai" and .path=="/v1/agents/sessions") or (.provider=="anthropic" and .path=="/v1/sessions") or (.provider=="gemini" and .path=="/v1beta/interactions") or (.provider=="xai" and .path=="/v1/responses")) | select(.method=="POST") | {provider, method, path, status, beta_header, verification}' generated/endpoints.json243```244245| Provider | Endpoint | Beta gate | Status | Live result | Required body | Notes |246|---|---|---|---|---|---|---|247| openai | `POST /v1/agents/sessions` | `OpenAI-Beta: agents=v1` | `DOCUMENTED` · `BETA` · `LIVE_VERIFIED` | success HTTP 201 | `environment` (required: `none` \| `openai_hosted` \| `self_hosted`), `agent` or `agent_id`, `input` (required when environment is `none`), `stream`, `vault_ids`, `metadata` | 201 `agent.session`; first turn starts immediately when `input` is given; send later input via `POST …/sessions/{id}/events` (`agent.session.input.message`); SSE with `stream:true` or `GET …/events?stream=true` |248| anthropic | `POST /v1/sessions` | `managed-agents-2026-04-01` | `DOCUMENTED` · `BETA` · `LIVE_VERIFIED` | success HTTP 200 | `agent` (id or `{type:agent,id,version}` or `agent_with_overrides`), `environment_id` (required), optional `initial_events[]` (≤50), `resources[]`, `vault_ids[]`, `budget`, `title`, `metadata` | 200 `session` (`sesn_…`); starts `idle` unless `initial_events`; send `user.message` via `POST /v1/sessions/{id}/events`; stream `GET /v1/sessions/{id}/events/stream`; agents and environments must exist first (`POST /v1/agents`, `POST /v1/environments`) |249| xai | `POST /v1/responses` | none (no beta headers on xAI) | `DOCUMENTED` · `LIVE_VERIFIED` | success HTTP 200 | `model`, `input`; agentic behaviour comes from `tools[]` (server-side `web_search`, `x_search`, `code_interpreter`, `file_search`, `mcp`, `image_generation` + your `function`/`shell` tools), `max_turns`, `store` (default true), `previous_response_id` | xAI has **no separate agent/session resource**: the Responses call itself runs the server-side agentic loop (bounded by `max_turns`) and the stored response (`store:true`, 30-day retention) is the session — continue with `previous_response_id`, inspect with `GET /v1/responses/{id}` / `…/input_items`, shrink with `POST /v1/responses/compact`. `grok-4.20-multi-agent-0309` (BETA) fans out to 4 or 16 agents inside one call. Grok Build is a CLI product, not an API |250| gemini | `POST /v1beta/interactions` | `/v1beta` path version | `DOCUMENTED` · `BETA` · `LIVE_VERIFIED` | success HTTP 200 | `model` **or** `agent` (`deep-research-preview-04-2026`, `deep-research-max-preview-04-2026`, `antigravity-preview-09-2026`, or a custom agent id from `POST /v1beta/agents`), `input`, optional `previous_interaction_id`, `store` (default true), `stream`, `background`, `tools[]`, `generation_config`, `agent_config`, `environment`, `webhook_config` | 200 `interaction` (`id`, `status`, `steps[]`/`outputs`); stateful chaining via `previous_interaction_id`; poll/resume with `GET /v1beta/interactions/{id}`, `POST …/cancel`, `DELETE`; long runs with `background:true` (Deep Research); sandboxed agents need `POST /v1beta/environments` (LIVE_VERIFIED) and can be scheduled with `/v1beta/triggers` (PREVIEW). `/v1/interactions` is documented GA but UNVERIFIED here. The **Live API** WebSocket (`BidiGenerateContent`, `setup` message) is the other session-shaped surface (voice) |251252Prerequisite creates: OpenAI `POST /v1/agents` (LIVE_VERIFIED 201) or an inline `agent`; Anthropic `POST /v1/agents` (LIVE_VERIFIED 200) **and** `POST /v1/environments` (LIVE_VERIFIED 200); Gemini `POST /v1beta/agents` (custom managed agents, PREVIEW, not tested) and `POST /v1beta/environments` (LIVE_VERIFIED). Not to be confused with: OpenAI **Conversations** (`POST /v1/conversations` — a state container for Responses, not an agent), **Realtime** client secrets (`POST /v1/realtime/client_secrets`), **ChatKit** sessions, the retired **Assistants** threads/runs; xAI `POST /v1/realtime/client_secrets` (voice session token); Gemini `POST /v1beta/auth_tokens` (ephemeral Live token) and `cachedContents` (a prefix cache, not a session). Full comparison: [agents-platforms](comparisons/agents-platforms.md).253254## Q6. Which feature needs a beta header (or a beta path / preview model)? <a id="q6"></a>255256The four providers gate pre-GA features differently: **Anthropic** by `anthropic-beta` header values (per parameter/tool/endpoint), **OpenAI** by `OpenAI-Beta` header values (per surface), **Gemini** by the **URL version** (`/v1beta` vs `/v1`) and `-preview` / `-exp` model ids (no header exists), **xAI** by nothing visible — no version or beta headers; gated features answer 403 ('alpha users') or 404 (ACL) and are recorded `ACCOUNT_RESTRICTED`.257258### Anthropic — `anthropic-beta` values that are still gating (status BETA / PREVIEW / DOCUMENTED-only), from `generated/fragments/headers/anthropic-beta-headers.json`, cross-referenced with the `beta_header` field of `generated/parameters.json`, `tools.json` and `endpoints.json`259260| Header value | Feature | Status | Models / scope | Gated parameters (sample) | Gated endpoints | Gated tool types |261|---|---|---|---|---|---|---|262| `computer-use-2025-01-24` | computer_20250124 tool version | `BETA` · `LEGACY` | claude-sonnet-4-5-20250929, claude-haiku-4-5-20251001, claude-opus-4-1-20250805 (retired), claude-sonnet-4-20250514 (retired) … | — | — | `computer_20250124` |263| `computer-use-2025-11-24` | computer_20251124 tool version (zoom, batch actions precursor) | `BETA` | claude-fable-5-1, claude-mythos-5-1, claude-fable-5, claude-mythos-5 … | POST /v1/messages → `tools[].enable_zoom` | — | `computer_20251124` |264| `mcp-client-2025-11-20` | MCP connector (mcp_toolset tool type) | `BETA` | — | POST /v1/messages → `mcp_servers[].authorization_token`; POST /v1/messages → `mcp_servers[].name`; POST /v1/messages → `mcp_servers[].tool_configuration` … (+5) | — | `mcp_toolset` |265| `dev-full-thinking-2025-05-14` | Full (unsummarized) thinking - developer/dev-only | `LIVE_DISCOVERED` | — | — | — | — |266| `interleaved-thinking-2025-05-14` | Interleaved thinking between tool calls with manual extended thinking | `BETA` | models using thinking type enabled (4.5 family, 4.6 manual mode); automatic (no header) with adaptive thinking; not supported on Haiku 4.5 | — | — | — |267| `context-management-2025-06-27` | Context editing (clear_tool_uses_20250919, clear_thinking_20251015 strategies) | `BETA` | all supported Claude models (context_management capability true on all 11 live ids) | POST /v1/messages → `context_management.edits[].clear_at_least`; POST /v1/messages → `context_management.edits[].clear_tool_inputs`; POST /v1/messages → `context_management.edits[].exclude_tools` … (+5) | — | — |268| `model-context-window-exceeded-2025-08-26` | Stop reason model_context_window_exceeded / request max possible tokens | `DOCUMENTED` | — | POST /v1/messages → `anthropic-beta: model-context-window-exceeded-2025-08-26` | — | — |269| `fast-mode-2026-02-01` | Fast mode (speed: "fast") | `PREVIEW` | claude-opus-5, claude-opus-4-8 | POST /v1/messages → `speed` | — | — |270| `output-300k-2026-03-24` | 300k max_tokens on the Message Batches API | `BETA` | claude-opus-5, claude-opus-4-8, claude-opus-4-7, claude-opus-4-6 … | — | — | — |271| `user-profiles-2026-03-24` | User profiles API (/v1/user_profiles) | `DOCUMENTED` · `BETA` | — | — | — | — |272| `user-profiles-2026-08-18` | User profiles API revision | `DOCUMENTED` · `BETA` | — | GET /v1/user_profiles → `limit`; GET /v1/user_profiles → `order_by`; GET /v1/user_profiles → `order` … (+30) | `GET /v1/user_profiles`; `GET /v1/user_profiles/{user_profile_id}`; `POST /v1/user_profiles` … (+2) | — |273| `user-profiles-2026-09-04` | User profiles API revision | `DOCUMENTED` · `BETA` | — | POST /v1/messages → `anthropic-user-profile-id`; POST /v1/messages/count_tokens → `anthropic-user-profile-id` | — | — |274| `advisor-tool-2026-03-01` | Advisor tool (advisor_20260301 server tool) | `BETA` | — | POST /v1/messages → `tools[].caching`; POST /v1/messages → `tools[].max_tokens`; POST /v1/messages → `tools[].model` | — | `advisor_20260301` |275| `managed-agents-2026-04-01` | Claude Managed Agents (agents, sessions, environments, deployments, vaults, skills in sessions, multiagent, outcomes, memory public beta 2026-04-23) | `BETA` | — | DELETE /v1/environments/{environment_id} → `environment_id`; DELETE /v1/sessions/{session_id} → `session_id`; DELETE /v1/sessions/{session_id}/resources/{resource_id} → `resource_id` … (+586) | `DELETE /v1/environments/{environment_id}`; `DELETE /v1/sessions/{session_id}`; `DELETE /v1/sessions/{session_id}/resources/{resource_id}` … (+59) | — |276| `cache-diagnosis-2026-04-07` | Cache diagnostics (diagnostics.previous_message_id -> cache_miss_reason) | `BETA` | — | POST /v1/messages → `diagnostics.previous_message_id`; POST /v1/messages → `diagnostics` | — | — |277| `dreaming-2026-04-21` | Dreams (memory store reorganization) for Managed Agents | `PREVIEW` | Fable 5, Sonnet 5 (2026-07-10), Opus 5 (2026-08-01) | GET /v1/dreams → `created_at[gt]`; GET /v1/dreams → `created_at[lt]`; GET /v1/dreams → `include_archived` … (+17) | `GET /v1/dreams`; `GET /v1/dreams/{dream_id}`; `POST /v1/dreams` … (+2) | — |278| `server-side-fallback-2026-06-01` | fallbacks parameter (re-run refused requests on another model) | `BETA` | — | — | — | — |279| `server-side-fallback-2026-07-01` | fallbacks with "default" mode (Anthropic-recommended fallback by refusal category) | `BETA` | — | POST /v1/messages → `fallbacks` | — | — |280| `fallback-credit-2026-06-01` | Fallback credit (billing credit when a fallback runs) | `BETA` | — | — | — | — |281| `fallback-credit-2026-07-01` | Fallback credit revision | `BETA` | — | POST /v1/messages → `fallback_credit_token` | — | — |282| `agent-memory-2026-07-22` | Memory stores list semantics (stable server order, depth 0/1, path_prefix segment match) | `BETA` | — | DELETE /v1/memory_stores/{memory_store_id} → `memory_store_id`; DELETE /v1/memory_stores/{memory_store_id}/memories/{memory_id} → `expected_content_sha256`; DELETE /v1/memory_stores/{memory_store_id}/memories/{memory_id} → `memory_id` … (+52) | `DELETE /v1/memory_stores/{memory_store_id}`; `DELETE /v1/memory_stores/{memory_store_id}/memories/{memory_id}`; `GET /v1/memory_stores` … (+11) | — |283| `mid-conversation-tool-changes-2026-07-01` | Add/remove tools mid-conversation (tool_addition / tool_removal blocks in role:system messages) preserving cache | `BETA` | claude-fable-5, claude-mythos-5, claude-opus-4-8, claude-opus-5 | — | — | — |284| `compact-2026-01-12` | Server-side compaction (context_management compact_20260112 strategy) | `BETA` | claude-fable-5-1, claude-mythos-5-1, claude-fable-5, claude-mythos-5 … | POST /v1/messages → `context_management.edits[].instructions`; POST /v1/messages → `context_management.edits[].pause_after_compaction` | — | — |285| `compact-2026-09-04` | Compact on demand (top-level compaction parameter -> signed compaction block) | `BETA` | — | POST /v1/messages → `compaction.instructions`; POST /v1/messages → `compaction.type`; POST /v1/messages → `compaction` | — | — |286| `mcp-tunnels-2026-06-22` | MCP tunnels API (/v1/tunnels) for private-network MCP servers | `PREVIEW` | — | GET /v1/tunnels → `include_archived`; GET /v1/tunnels → `limit`; GET /v1/tunnels → `page` … (+16) | `GET /v1/tunnels`; `GET /v1/tunnels/{tunnel_id}`; `GET /v1/tunnels/{tunnel_id}/certificates` … (+7) | — |287| `task-budgets-2026-03-13` | Task budgets (advisory token budget for a whole agentic loop) | `BETA` | claude-fable-5-1, claude-mythos-5-1, claude-fable-5, claude-mythos-5 … | POST /v1/messages → `output_config.task_budget.remaining`; POST /v1/messages → `output_config.task_budget.total`; POST /v1/messages → `output_config.task_budget.type` … (+1) | — | — |288| `thinking-display-updates-2026-08-18` | thinking.display: "updates" (progress updates between tool calls as text) | `BETA` | claude-fable-5-1, claude-mythos-5-1, claude-fable-5 | — | — | — |289| `mid-conversation-output-config-2026-07-01` | Per-message effort (output_config.effort on role:system messages) | `BETA` | claude-fable-5-1, claude-mythos-5-1, claude-opus-5 | POST /v1/messages → `messages[].output_config.effort` | — | — |290| `thinking-binding-controls-2026-08-01` | Preserved-thinking binding: input_transformations response field + thinking.block_binding.prefix_mismatch_behavior | `BETA` | Fable 5.1 / Mythos 5.1 (prefix check enforced for accounts created on/after 2026-08-31) | POST /v1/messages → `thinking.block_binding.prefix_mismatch_behavior`; POST /v1/messages → `thinking.block_binding` | — | — |291| `mid-conversation-system-clear-at-2026-08-21` | Turn-scoped system messages (clear_at: "next_user_message") | `BETA` | — | — | — | — |292293Graduated (header now optional — `LEGACY`): `message-batches-2024-09-24`, `prompt-caching-2024-07-31`, `pdfs-2024-09-25`, `token-counting-2024-11-01`, `computer-use-2025-01-24`, `token-efficient-tools-2025-02-19`, `output-128k-2025-02-19`, `files-api-2025-04-14`, `code-execution-2025-05-22`, `extended-cache-ttl-2025-04-11`, `skills-2025-10-02`, `thinking-token-count-2026-05-13`, `structured-outputs-2025-11-13`, `ce-user-management-2026-07-13`, `fine-grained-tool-streaming-2025-05-14`, `search-results-2025-06-09`, `effort-2025-11-24`. Retired / deprecated: `computer-use-2024-10-22`, `mcp-client-2025-04-04`, `context-1m-2025-08-07`, `max-tokens-3-5-sonnet-2024-07-15`.294295Live findings worth knowing: `mcp_servers` without a header → 400 that names `mcp-client-2026-09-15`, which is itself rejected (use `mcp-client-2025-11-20`); memory-store calls reject the pair `agent-memory-2026-07-22` + `managed-agents-2026-04-01`; `speed` and `context_management` are rejected with 400 'Extra inputs are not permitted' without their headers; `output_format` (legacy) → 400 even with `structured-outputs-2025-11-13`.296297### OpenAI — `OpenAI-Beta` values (from `generated/endpoints.json` / `headers.json`)298299| Header value | Surface | Endpoints | Status |300|---|---|---|---|301| `OpenAI-Beta: agents=v1` | agents-platform/agents | 43 (e.g. `GET /v1/agents/environments/{environment_id}`) | `BETA` · `DOCUMENTED` · `LIVE_DISCOVERED` · `LIVE_VERIFIED` |302| `OpenAI-Beta: assistants=v2 (historical)` | agents-platform/assistants | 23 (e.g. `GET /v1/assistants`) | `DOCUMENTED` · `RETIRED` |303| `OpenAI-Beta: chatkit_beta=v1` | agents-platform/chatkit | 6 (e.g. `POST /v1/chatkit/sessions/{session_id}/cancel`) | `BETA` · `DOCUMENTED` · `LIVE_DISCOVERED` · `LIVE_VERIFIED` |304| `OpenAI-Beta: realtime=v1` | realtime | 2 (e.g. `POST /v1/realtime/sessions`) | `BETA` · `DOCUMENTED` · `FAILED_VERIFICATION` · `LEGACY` |305| `OpenAI-Beta: workspace_agent_runs=v1 (required to get agent_trigger_run_id / use runs polling)` | agents-platform/workspace-agents | 2 (e.g. `POST /v1/workspace_agents/{id}/trigger`) | `BETA` · `DOCUMENTED` · `UNVERIFIED` |306| `openai-beta (optional, array of strings: "Optional beta features to enable for this request")` | responses | 1 (e.g. `POST /v1/responses?beta=true`) | `BETA` · `DOCUMENTED` · `LIVE_VERIFIED` |307308Everything else on OpenAI (Responses, hosted tools incl. MCP, structured outputs, prompt caching, Batch, Files, Webhooks, Realtime GA) needs **no** beta header; `POST /v1/responses?beta=true` exposes a beta schema (`multi_agent`) with an optional body-level `openai-beta[]`. `OpenAI-Beta: realtime=v1` is legacy (the beta endpoints return 404).309310### Gemini — `/v1beta` vs `/v1`, `-preview` models, BETA/PREVIEW endpoints (from `generated/endpoints.json`, `models.json`)311312```bash313# Gemini surfaces that exist only under /v1beta314jq -r '[.[] | select(.provider=="gemini" and (.path|startswith("/v1beta/")))] | group_by(.api_family) | .[] | "\(.[0].api_family)\t\(length)"' generated/endpoints.json315# preview / experimental Gemini model ids still served316jq -r '.[] | select(.provider=="gemini" and (.kind=="preview" or .kind=="experimental") and (.status|index("RETIRED")|not)) | .id' generated/models.json317```318319- **Stable `/v1` surface** (9 paths recorded): `/v1/interactions`, `/v1/interactions/{id}`, `/v1/interactions/{id}/cancel`, `/v1/models`, `/v1/models/{model}:generateContent`, `/v1/webhooks`, `/v1/webhooks/{id}`, `/v1/webhooks/{id}/rotate_secret`, `/v1/webhooks/{id}:ping`. Everything else — including **explicit context caching (`cachedContents`), Files, File Search stores, Batch, embeddings, `countTokens`, Live API (`v1beta` WebSocket), ephemeral `auth_tokens`, Veo `predictLongRunning`, the OpenAI-compatibility layer and the managed-agents preview (`agents`, `environments`, `credentials`, `triggers`)** — is **v1beta-only** (families: `agents`, `auth-tokens`, `batch`, `caching`, `credentials`, `dynamic`, `embeddings`, `environments`, `file-search`, `files`, `image-generation`, `legacy-palm`, `openai-compat`, `operations`, `tokens`, `transcription`, `triggers`, `tuning`, `video-generation`). `/v1/cachedContents` returns 404.320- **Endpoints recorded BETA/PREVIEW** (64): `agents` (4), `auth-tokens` (1), `caching` (5), `credentials` (5), `dynamic` (2), `environments` (7), `file-search` (12), `interactions` (4), `live` (2), `music-generation` (1), `openai-compat` (7), `triggers` (7), `video-generation` (1), `webhooks` (6). `POST /v1beta/interactions` is BETA + LIVE_VERIFIED; its `/v1/interactions` twin is documented GA but UNVERIFIED.321- **Preview / experimental model ids still served** (25): `gemini-2.5-flash-preview-tts` (PREVIEW), `gemini-2.5-pro-preview-tts` (PREVIEW), `gemini-3-flash-preview` (PREVIEW), `gemini-3.1-pro-preview` (PREVIEW/ACCOUNT_RESTRICTED), `gemini-3.1-pro-preview-customtools` (PREVIEW), `gemini-omni-flash-preview` (PREVIEW/DEPRECATED), `lyria-3-clip-preview` (PREVIEW), `lyria-3-pro-preview` (PREVIEW), `gemini-3.1-flash-tts-preview` (PREVIEW), `gemini-robotics-er-2-preview` (PREVIEW), `gemini-2.5-computer-use-preview-10-2025` (PREVIEW), `gemini-embedding-2-preview` (PREVIEW), `veo-3.1-generate-preview` (PREVIEW), `veo-3.1-fast-generate-preview` (PREVIEW), `veo-3.1-lite-generate-preview` (PREVIEW), `gemini-2.5-flash-native-audio-preview-09-2025` (PREVIEW), `gemini-2.5-flash-native-audio-preview-12-2025` (PREVIEW), `gemini-3.1-flash-live-preview` (PREVIEW), `gemini-robotics-er-2-streaming-preview` (PREVIEW), `gemini-3.5-live-translate-preview` (PREVIEW), `lyria-realtime-exp` (BETA), `gemini-2.0-flash-exp` (BETA/UNVERIFIED), `gemini-2.5-flash-preview-09-2025` (PREVIEW/UNVERIFIED), `lyria-3.5-clip-preview` (PREVIEW/UNVERIFIED), `lyria-3.5-pro-preview` (PREVIEW/UNVERIFIED). Preview models 'may be used in production' but carry more restrictive limits and can be deprecated with two weeks' notice; `-latest` aliases (`gemini-flash-latest`, `gemini-pro-latest`, `gemini-flash-lite-latest`) are hot-swapped.322- **Gated by account tier rather than header**: Pro models and paid-only media models return 429 `limit: 0` on the free tier (`gemini-3.1-pro-preview`, `gemini-pro-latest`, `cachedContents` create — all `ACCOUNT_RESTRICTED` here); Gemini 2.5 models return 404 'no longer available to new users'.323324### xAI — no beta headers; gating is by ACL / alpha program (from `generated/endpoints.json`, `tools.json`, `headers.json`)325326- `headers.json` records **no versioning / beta headers** for xAI: features are selected by body fields (`service_tier`, `reasoning_effort`, `store`, `prompt_cache_key`, `include`) or by base URL (`us.api.x.ai`, `management-api.x.ai`).327- **ACCOUNT_RESTRICTED endpoints** (41): `collections` (11), `embeddings` (1), `management` (21), `models` (2), `skills` (5), `voice` (1) — the Management API needs a separate Management key (401 code 16 with an inference key), `/v1/embeddings` and `grok-embedding-small` return 404 for this team, `/v1/skills` returns 404, custom-voice creation is Enterprise-only, and Collections are documented on `management-api.x.ai` although the same paths answered on `api.x.ai` (LIVE_DISCOVERED).328- **Tools**: `tool_search` → `DOCUMENTED` · `ACCOUNT_RESTRICTED`; `live_search` → `DOCUMENTED` · `RETIRED`. `tool_search` (+ `defer_loading`) answers 403 'only available for alpha users'; `live_search` / `search_parameters` / `web_search_options` answer 410.329- **Models flagged BETA/PREVIEW**: `grok-4.20-multi-agent-0309` (BETA), `grok-build-0.1` (PREVIEW) (multi-agent is Responses-only; Grok Build is 'early access'). Regional host `eu-west-1.api.x.ai` is undocumented (LIVE_DISCOVERED).330331## Q7. Which model supports combination Z? <a id="q7"></a>332333Recipe: filter `generated/models.json` on `capabilities.*` booleans. Key families — **OpenAI**: `image_in`, `reasoning`, `structured_outputs`, `function_calling`, `prompt_caching`, `tool_web_search`, `tool_code_interpreter`, `tool_mcp`, `tool_computer_use`, `tool_hosted_shell`, `tool_apply_patch`, `tool_skills`, `tool_tool_search`, `fine_tuning`, `batch`, `audio_in/out`, `realtime`; **Anthropic**: `vision`, `pdf_input`, `structured_outputs_json`, `strict_tool_use`, `prompt_caching`, `web_search`, `web_fetch`, `code_execution`, `programmatic_tool_calling`, `computer_use_toolset_ga`, `browser_use`, `mcp_connector`, `agent_skills`, `tool_search`, `effort_parameter`, `thinking_adaptive`, `context_1m_default_no_beta`, `fast_mode_speed_fast`, `batch_api`, `zero_data_retention_eligible`; **xAI**: `reasoning`, `reasoning_can_be_disabled`, `reasoning_effort_levels[]`, `function_calling`, `structured_outputs`, `prompt_caching_automatic`, `batch_api`, `priority_processing`, `context_compaction`, `websocket_responses`, `web_search`, `x_search`, `code_execution`, `collections_search`, `remote_mcp`, `image_generation_tool`, `files_attachments`, `encrypted_reasoning_content`, `anthropic_messages_compat`; **Gemini**: `thinking`, `thinking_level_param`, `thinking_levels[]`, `thought_signatures`, `structured_output`, `function_calling`, `parallel_function_calling`, `compositional_function_calling`, `google_search_grounding`, `google_maps_grounding`, `url_context`, `code_execution`, `computer_use`, `file_search`, `context_caching_explicit`, `context_caching_implicit`, `batch_api`, `flex_inference`, `priority_inference`, `live_api`, `interactions_api`, `openai_compatible_chat`, `audio_input`, `video_input`, `pdf_input`, `image_output`, `audio_output`, `embeddings`, `tuning`. Flat matrices: `generated/compatibility/model-capability-matrix.json` (+ `.csv`), `gemini-feature-model-matrix.json`, `gemini-tool-model-matrix.json`.334335```bash336# generic pattern337jq -r '.[] | select(.provider=="anthropic" and .capabilities.context_1m_default_no_beta==true and .capabilities.computer_use_toolset_ga==true and .capabilities.web_search==true and (.status|index("RETIRED")|not)) | .id' generated/models.json338jq -r '.[] | select(.provider=="gemini" and .capabilities.live_api==true and .capabilities.function_calling==true and .capabilities.google_search_grounding==true and (.status|index("RETIRED")|not)) | .id' generated/models.json339```340341| Combination | Provider | Matching models (status) |342|---|---|---|343| Z1 (OpenAI): ≥1M context + structured outputs + prompt caching + web search + code interpreter + computer use | openai | `gpt-5.4`, `gpt-5.4-2026-03-05`, `gpt-5.5`, `gpt-5.5-2026-04-23`, `gpt-5.6`, `gpt-5.6-luna`, `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-6-astra`, `gpt-daybreak-blue-latest` (ACCOUNT_RESTRICTED) |344| Z1 (Anthropic): 1M context (no header) + structured outputs + web search + code execution + computer-use toolset GA | anthropic | `claude-fable-5-1`, `claude-mythos-5-1` (ACCOUNT_RESTRICTED/PREVIEW), `claude-fable-5` (LEGACY), `claude-mythos-5` (ACCOUNT_RESTRICTED/PREVIEW), `claude-opus-5`, `claude-opus-4-8` (LEGACY), `claude-sonnet-5` |345| Z1 (xAI): ≥1M context + structured outputs + automatic prompt caching + web_search + code_execution + batch | xai | — none |346| Z1 (Gemini): ≥1M context + structured output + explicit context caching + Google Search grounding + code execution + computer use | gemini | `gemini-3-flash-preview` (PREVIEW), `gemini-3.5-flash`, `gemini-3.5-flash-lite`, `gemini-3.6-flash`, `gemini-3.7-flash`, `gemini-3.8-flash` |347| Z2 (OpenAI): reasoning can be switched off (`effort: none` accepted) + function calling + image input | openai | `gpt-5.1`, `gpt-5.1-2025-11-13`, `gpt-5.2`, `gpt-5.2-2025-12-11`, `gpt-5.4`, `gpt-5.4-2026-03-05`, `gpt-5.4-mini`, `gpt-5.4-mini-2026-03-17`, `gpt-5.4-nano`, `gpt-5.4-nano-2026-03-17`, `gpt-5.5`, `gpt-5.5-2026-04-23`, `gpt-5.6`, `gpt-5.6-luna`, `gpt-5.6-sol`, `gpt-5.6-terra` |348| Z2 (Anthropic): thinking can be disabled + effort parameter + tool use | anthropic | `claude-mythos-preview` (ACCOUNT_RESTRICTED/DEPRECATED/PREVIEW), `claude-opus-5`, `claude-opus-4-8` (LEGACY), `claude-opus-4-7` (LEGACY), `claude-opus-4-6` (LEGACY), `claude-opus-4-5-20251101` (LEGACY), `claude-sonnet-5`, `claude-sonnet-4-6` (LEGACY) |349| Z2 (xAI): reasoning can be disabled (`reasoning_effort: none`) or is absent + function calling + image input | xai | `grok-4.3`, `grok-4.20-0309-non-reasoning` |350| Z2 (Gemini): `thinkingLevel` parameter with `minimal` accepted + function calling + image input | gemini | `gemini-3-flash-preview` (PREVIEW), `gemini-3.1-flash-lite` (DEPRECATED), `gemini-3.5-flash`, `gemini-3.5-flash-lite`, `gemini-3.6-flash` |351| Z3 (OpenAI): fine-tuning supported (self-serve winding down, 2027-01-06) | openai | `babbage-002` (DEPRECATED/LEGACY), `davinci-002` (DEPRECATED/LEGACY), `gpt-3.5-turbo` (DEPRECATED/LEGACY), `gpt-3.5-turbo-0125` (DEPRECATED/LEGACY), `gpt-3.5-turbo-1106` (DEPRECATED/LEGACY), `gpt-3.5-turbo-instruct` (DEPRECATED/LEGACY), `gpt-4` (DEPRECATED/LEGACY), `gpt-4-0613` (DEPRECATED/LEGACY), `gpt-4.1`, `gpt-4.1-2025-04-14`, `gpt-4.1-mini`, `gpt-4.1-mini-2025-04-14`, `gpt-4.1-nano` (DEPRECATED), `gpt-4.1-nano-2025-04-14` (DEPRECATED), `gpt-4o`, `gpt-4o-2024-05-13` (DEPRECATED/LEGACY), `gpt-4o-2024-08-06`, `gpt-4o-2024-11-20`, `gpt-4o-mini`, `gpt-4o-mini-2024-07-18`, `o4-mini` (DEPRECATED), `o4-mini-2025-04-16` (DEPRECATED) |352| Z3 (Gemini): tuning supported on the Developer API | gemini | — none |353| Z4 (OpenAI): audio in + audio out (chat or realtime) | openai | `gpt-4o-audio-preview-2024-12-17` (PREVIEW), `gpt-4o-audio-preview-2025-06-03` (PREVIEW), `gpt-4o-mini-audio-preview-2024-12-17` (PREVIEW), `gpt-4o-mini-realtime-preview-2024-12-17` (PREVIEW), `gpt-audio` (DEPRECATED), `gpt-audio-1.5`, `gpt-audio-2025-08-28`, `gpt-audio-mini` (DEPRECATED), `gpt-audio-mini-2025-12-15`, `gpt-live-1`, `gpt-realtime` (DEPRECATED), `gpt-realtime-1.5`, `gpt-realtime-2`, `gpt-realtime-2.1`, `gpt-realtime-2.1-mini`, `gpt-realtime-2025-08-28`, `gpt-realtime-mini` (DEPRECATED), `gpt-realtime-mini-2025-12-15`, `gpt-realtime-translate` |354| Z4 (Anthropic): `speed: fast` supported | anthropic | `claude-opus-5`, `claude-opus-4-8` (LEGACY) |355| Z4 (xAI): speech-to-speech + function calling + web/x search in the voice session | xai | `grok-voice-think-fast-2.0` |356| Z4 (Gemini): Live API (audio in + audio out) + function calling + Google Search | gemini | `gemini-2.5-flash-native-audio-preview-12-2025` (PREVIEW), `gemini-3.1-flash-live-preview` (PREVIEW), `gemini-3.8-live` |357| Z5 (OpenAI): flex tier + batch + 24h extended cache retention documented | openai | `gpt-5`, `gpt-5-2025-08-07` (DEPRECATED), `gpt-5.1`, `gpt-5.1-2025-11-13`, `gpt-5.2`, `gpt-5.2-2025-12-11`, `gpt-5.4`, `gpt-5.4-2026-03-05`, `gpt-5.5`, `gpt-5.5-2026-04-23`, `gpt-5.5-pro`, `gpt-5.5-pro-2026-04-23`, `gpt-5.6`, `gpt-5.6-luna`, `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-6-astra` |358| Z5 (Anthropic): zero-data-retention eligible + programmatic tool calling + MCP connector | anthropic | `claude-mythos-preview` (ACCOUNT_RESTRICTED/DEPRECATED/PREVIEW), `claude-opus-5`, `claude-opus-4-8` (LEGACY), `claude-opus-4-7` (LEGACY), `claude-opus-4-6` (LEGACY), `claude-opus-4-5-20251101` (LEGACY), `claude-sonnet-5`, `claude-sonnet-4-6` (LEGACY), `claude-sonnet-4-5-20250929` (LEGACY) |359| Z5 (xAI): priority processing + WebSocket Responses + context compaction + remote MCP | xai | — none |360| Z5 (Gemini): flex + priority + batch tiers + implicit caching | gemini | `gemini-2.5-flash`, `gemini-2.5-pro`, `gemini-2.5-flash-lite` (ACCOUNT_RESTRICTED), `gemini-3-flash-preview` (PREVIEW), `gemini-3.1-pro-preview` (PREVIEW/ACCOUNT_RESTRICTED), `gemini-3.1-pro-preview-customtools` (PREVIEW), `gemini-3.1-flash-lite` (DEPRECATED), `gemini-3.5-flash`, `gemini-3.5-flash-lite`, `gemini-3.6-flash`, `gemini-3.7-flash`, `gemini-3.8-flash` |361| Z6 (OpenAI): skills + hosted shell + apply_patch + tool search (full coding toolset) | openai | `gpt-5.4`, `gpt-5.4-2026-03-05`, `gpt-5.4-mini`, `gpt-5.4-mini-2026-03-17`, `gpt-5.5`, `gpt-5.5-2026-04-23`, `gpt-5.6`, `gpt-5.6-cyber` (ACCOUNT_RESTRICTED), `gpt-5.6-luna`, `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-6-astra`, `gpt-daybreak-blue-latest` (ACCOUNT_RESTRICTED), `gpt-daybreak-red-latest` |362| Z6 (Anthropic): per-message effort (beta) + task budgets (beta) + adaptive thinking | anthropic | `claude-fable-5-1`, `claude-mythos-5-1` (ACCOUNT_RESTRICTED/PREVIEW), `claude-opus-5` |363| Z6 (xAI): x_search + collections_search + image_generation tool + files attachments (full Grok agent toolset) | xai | — none |364| Z6 (Gemini): Google Maps grounding + URL context + File Search + thought signatures (full Gemini 3 toolset) | gemini | `gemini-3-flash-preview` (PREVIEW), `gemini-3.1-pro-preview` (PREVIEW/ACCOUNT_RESTRICTED), `gemini-3.1-pro-preview-customtools` (PREVIEW), `gemini-3.1-flash-lite` (DEPRECATED), `gemini-3.5-flash`, `gemini-3.5-flash-lite`, `gemini-3.6-flash`, `gemini-3.7-flash`, `gemini-3.8-flash` |365| Z7 (Gemini): free tier + Google Search grounding + structured output (zero-cost prototyping) | gemini | `gemini-2.5-flash`, `gemini-2.5-pro`, `gemini-2.5-flash-lite` (ACCOUNT_RESTRICTED), `gemini-3-flash-preview` (PREVIEW), `gemini-3.1-flash-lite` (DEPRECATED), `gemini-3.5-flash`, `gemini-3.5-flash-lite`, `gemini-3.6-flash`, `gemini-3.7-flash`, `gemini-3.8-flash`, `gemini-robotics-er-2-preview` (PREVIEW) |366| Z7 (xAI): OpenAI-compatible chat + Anthropic-compatible `/v1/messages` + deferred completions | xai | `grok-4.6`, `grok-4.5`, `grok-4.3`, `grok-4.20-0309-reasoning`, `grok-4.20-0309-non-reasoning`, `grok-build-0.1` (PREVIEW) |367368Snapshot ids (e.g. `gpt-5.4-2026-03-05`) share their alias's capabilities and appear in the raw query output; the table above keeps them because they are distinct callable ids. Capability flags marked `"unknown"` in the source never match a `== true` filter — check the model page before excluding a model on that basis (many Gemini media/agent records and Gemma 4 carry `"unknown"` for tool flags). Gemini `computer_use` / `file_search` may be the string `"Supported (Preview)"` / `"Supported (AI Studio only)"` — treated as true above (✔*).369370## Q8. Which provider offers X, and at what price? <a id="q8"></a>371372Computed from `generated/pricing.json` (tool/service rows) and the `pricing` blocks of `generated/models.json`. Units are quoted as recorded; '—' = no documented surface. Statuses are those of the underlying tool/model record.373374```bash375# every per-call tool price across providers376jq -r '.[] | select(.unit|test("call|search|request|prompt";"i")) | [.provider, .model_or_service, .dimension, (.price|tostring), .unit, (.tier//"")] | @tsv' generated/pricing.json377```378379| Capability | OpenAI | Anthropic | xAI | Gemini |380|---|---|---|---|---|381| Web search (per 1k) | $10 per 1k calls (`web_search`; preview $25/1k on non-reasoning models) | $10 per 1,000 searches (`web_search_2026*`; failed searches free) | $5 per 1K calls (`web_search`, image search included) | Gemini 3.x: $14 per 1K requests after 5,000 free/month (billed per **query**); Gemini 2.5: $35 per 1K grounded prompts after 1,500 RPD free; tool `googleSearch` ACCOUNT_RESTRICTED on this key |382| Social / maps search | — | — | `x_search` $5 per 1K calls until 2026-09-21, then $5 per 1K posts + $10 per 1K profiles | `googleMaps` $14 per 1K search queries (Gemini 3.x, 5,000 free/month); $25 / 1k grounded prompts on 2.5 |383| URL fetch / context | — (web search opens pages) | `web_fetch` $0 per fetch (content tokens) | — (`open_page` action of web search) | `urlContext` $0 per call + content billed as input tokens |384| Code execution | $0.03 per 20-minute session per container (by size) (1 GB; 4 GB $0.12, 16 GB $0.48, 64 GB $1.92; per-minute billing since 2026-06-02) | $0.05 per hour per container after 1,550 free container-hours/org/month; free alongside web_search/web_fetch 20260209+ | $5 per 1K calls (`code_interpreter` / `code_execution`) | $0 per call — `codeExecution` has no fee; code + results billed as tokens |385| Managed RAG / file search | $2.5 per 1k calls + storage $0.1 per GB per day (1 GB free) | — (you retrieve; `search_result` blocks) | `file_search` / `collections_search` $2.5 per 1K calls + storage $0.1 per GiB per day; implicit `attachment_search` $10 per 1K calls | `fileSearch` indexing $0.15 per 1M tokens once; storage and queries free; retrieved chunks billed as input |386| Image generation (per image) | gpt-image-2: $0.006 per image (low 1024²) … $0.211 per image (high 1024²); token rates $5 text in / $8 image in / $30 image out per 1M | — | grok-imagine-image $0.02; grok-imagine-image-2.0 $0.04–$0.08 (quality × resolution); grok-imagine-image-quality $0.05 (DEPRECATED) | gemini-3.1-flash-lite-image $0.0336 (1K); gemini-3.1-flash-image $0.045–$0.151 (0.5K–4K); gemini-3-pro-image $0.134 (1K/2K) / $0.24 (4K); batch 50 % |387| Realtime voice (audio minutes) | gpt-realtime-2.1 audio $32 per 1M tokens in / $64 per 1M tokens out (≈ $0.06 in + $0.24 out per minute at ~10 tok/s... token-based); gpt-live-1 $0.05 per minute (billed per second) | — | grok-voice-think-fast-2.0 **$0.08 / min** ($4.8 / h) audio sent or received + $0.004 per text item | gemini-3.8-live audio in $3.0 / 1M (≈ $0.005 / min), audio out $12.0 / 1M (≈ $0.018 / min); free tier available |388| Text-to-speech | gpt-4o-mini-tts $0.6 per 1M tokens text in + $12 per 1M tokens audio out; tts-1 $15 / 1M chars | — | $15 per 1M characters (`POST /v1/tts`, 28 voices, `language` required) | gemini-3.1-flash-tts-preview $1.0 text in + $20.0 audio out per 1M (25 tok/s ≈ $0.030 / min); free tier |389| Speech-to-text | gpt-transcribe $0.0045 per minute; whisper-1 $0.006 / min (→ 2027-02-26) | — | $0.1 per hour REST, $0.2 per hour streaming (grok-voice-transcribe-2.0 / 1.0) | gemini-3.5-transcribe audio in $2.0 / 1M (≈ $0.003 / min) + text out $12.0 / 1M (blended blended ~$0.005/min); free tier |390| Embeddings (per 1M tokens) | text-embedding-3-small $0.02 per 1M tokens, -large $0.13 per 1M tokens | — | `grok-embedding-small`: no published price (ACCOUNT_RESTRICTED, 404) | gemini-embedding-2 text $0.2 (batch $0.1); image $0.45, audio $6.5, video $12.0; free tier |391| Video generation (per second) | sora-2 $0.10 (720p), sora-2-pro $0.30–$0.70 — **shutdown 2026-09-24** | — | grok-imagine-video $0.05; grok-imagine-video-1.5 $0.08 | Veo 3.1 $0.4 (720p/1080p) / $0.6 (4K); Veo 3.1 Fast $0.1 / $0.12 / $0.3; Veo 3.1 Lite $0.05 / $0.08 |392| Music generation | — | — | — | lyria-3.5 $0.08 per song (full length); lyria-3-clip-preview $0.04 per song (30 s clip) |393| Token counting | free (`POST /v1/responses/input_tokens`) | free (`POST /v1/messages/count_tokens`) | free (`POST /v1/tokenize-text`) | free (`:countTokens`) |394395Provider-specific billing quirks that change the arithmetic: xAI bills reasoning tokens on every Grok call and applies its **≥200k-token long-context rate to the whole request**; Gemini's 3.6–3.8 Flash prices double on 2027-01-01 and its free tier covers most Flash/Live/TTS/embedding models; Anthropic and OpenAI charge cache **writes** (1.25×) on their newest models while xAI and Gemini implicit caching have none (Gemini explicit caches pay storage per token-hour). Full tables and cost models: [pricing comparison](comparisons/pricing.md).396397## Q9. Where do I look for … ? <a id="q9"></a>398399| Question | File(s) |400|---|---|401| Every endpoint, with auth/beta/verification | `generated/endpoints.json` (870 records) · [docs/endpoints/index.md](endpoints/index.md) · [by-status](endpoints/by-status.md) |402| Every parameter of an endpoint | `generated/parameters.json` (7,862 rows) filtered by `endpoint` |403| Model facts (context, output, cutoff, prices, capabilities, cloud ids) | `generated/models.json` (396 records) · [docs/comparisons/models.md](comparisons/models.md) · docs/models/{openai,anthropic,xai,gemini}-models.md |404| Tool definitions & result shapes | `generated/tools.json` (69 records) · docs/tools/{openai,anthropic,xai,gemini}/*.md |405| Prices | `generated/pricing.json` (1,522 rows) · [docs/comparisons/pricing.md](comparisons/pricing.md) · docs/{openai,anthropic,xai,gemini}/pricing.md |406| Errors and retry rules | `generated/errors.json` · docs/errors/{openai,anthropic,gemini}.md · docs/xai/authentication-headers-errors.md |407| Headers (request/response/webhook) | `generated/headers.json` · `generated/fragments/headers/{anthropic-beta-headers,anthropic-headers,openai-headers,xai-headers,gemini-headers}.json` |408| Streaming & webhook events | `generated/streaming-events.json` (506) · `generated/webhook-events.json` (72; OpenAI + Anthropic only — xAI has the single SIP `realtime.call.incoming`, Gemini Interactions use `webhook_config` / `/v1/webhooks`) |409| Deprecations / retirements | `generated/deprecations.json` · docs/openai/deprecations.md · docs/anthropic/deprecations.md · docs/xai/deprecations-and-release-notes.md · docs/gemini/deprecations-and-changelog.md |410| Feature-by-feature comparison (4 providers) | [docs/comparisons/features.md](comparisons/features.md) · `generated/compatibility/cross-provider-feature-matrix.json` |411| Rate limits | `generated/rate-limits.json` · docs/openai/rate-limits.md · docs/anthropic/rate-limits.md · docs/xai/rate-limits.md · docs/gemini/rate-limits.md |412| SDKs | `generated/sdks.json` · docs/openai/sdks.md · docs/anthropic/sdks.md · docs/xai/sdks.md · docs/gemini/sdks.md |413| Cloud platform availability | `generated/compatibility/anthropic-feature-platform-matrix.json` · docs/anthropic/cloud-providers.md · docs/openai/data-residency-and-regions.md · docs/gemini/vertex-vs-gemini-api.md · docs/xai/sdks.md (Vertex Model Garden / Foundry) |414| OpenAI-compatibility layers | docs/xai/chat-completions.md · docs/xai/messages-compat.md · docs/gemini/openai-compatibility.md |415| Runnable examples and their status | `generated/examples-manifest.json` · `examples/` |416| Live calls made by this atlas (cost, status) | `reports/live-requests.jsonl` |