{
  "experiment": "expH_capture_cost_frontier",
  "run": 3,
  "scope": "capture overhead on a real 4-bit checkpoint (mlx-lm prefill)",
  "commit": "f3dd60ef6bed19c96456381481abb0aa50b83b73",
  "config": {
    "model": "mlx-community/Qwen3-0.6B-4bit",
    "seq": 901,
    "n_layers": 28,
    "d_model": 1024,
    "n_forwards": 10,
    "warmup": 3,
    "repeats": 3,
    "seed": 0
  },
  "manifest": {
    "author": "Simon-Pierre Boucher",
    "contact": "contact@spboucher.ai",
    "website": "https://modelmap.io",
    "chip": {
      "brand": "Apple M5 Max",
      "cores_total": 18,
      "cores_performance": 6,
      "cores_efficiency": 12
    },
    "memory": {
      "unified_gb": 48.0,
      "pagesize": 16384
    },
    "os": {
      "system": "Darwin",
      "version": "27.0",
      "arch": "arm64"
    },
    "software": {
      "python": "3.14.4",
      "numpy": "2.5.2",
      "mlx": "0.32.0",
      "torch": "2.13.0",
      "safetensors": "0.8.0"
    }
  },
  "compute": [
    {
      "backend": "mlx-lm",
      "mode": "plain",
      "s_per_forward": [
        0.02832533750006405,
        0.028470029100026294,
        0.02862457089995587
      ],
      "tokens_per_s_mean": 31643.66632788203
    },
    {
      "backend": "mlx-lm",
      "mode": "retain",
      "s_per_forward": [
        0.028357170900017083,
        0.028654929099866423,
        0.028654850000020816
      ],
      "tokens_per_s_mean": 31552.424826645736
    },
    {
      "backend": "mlx-lm",
      "mode": "retain+copy+write",
      "s_per_forward": [
        0.03822784169988154,
        0.0356740749999517,
        0.03538348749989382
      ],
      "tokens_per_s_mean": 24733.40351159859
    }
  ]
}