{
  "gpu": "NVIDIA H100 80GB HBM3",
  "torch": "2.8.0+cu129",
  "results": [
    {
      "mb": 8,
      "ctx": 1024,
      "compile": "default",
      "tok_s": 112855.93922615414,
      "peak_gb": 18.579795968,
      "tflops": 339.30627471695726,
      "warmup_s": 26.683147192001343
    },
    {
      "mb": 12,
      "ctx": 1024,
      "compile": "default",
      "tok_s": 118795.12738506112,
      "peak_gb": 24.264536064,
      "tflops": 357.1627014399097,
      "warmup_s": 47.67642426490784
    },
    {
      "mb": 16,
      "ctx": 1024,
      "compile": "default",
      "tok_s": 122837.53238217832,
      "peak_gb": 29.92830464,
      "tflops": 369.31636734242323,
      "warmup_s": 0.4134063720703125
    },
    {
      "mb": 24,
      "ctx": 1024,
      "compile": "default",
      "tok_s": 125688.80943331693,
      "peak_gb": 41.244962816,
      "tflops": 377.8888554280444,
      "warmup_s": 0.5790824890136719
    },
    {
      "mb": 32,
      "ctx": 1024,
      "compile": "default",
      "tok_s": 127464.49062053277,
      "peak_gb": 52.637249536,
      "tflops": 383.22751791094504,
      "warmup_s": 0.75799560546875
    },
    {
      "mb": 48,
      "ctx": 1024,
      "compile": "default",
      "tok_s": 130583.87669116902,
      "peak_gb": 75.252740096,
      "tflops": 392.606087388893,
      "warmup_s": 1.1335768699645996
    },
    {
      "mb": 8,
      "ctx": 2048,
      "compile": "default",
      "tok_s": 111115.13829731975,
      "peak_gb": 29.924896768,
      "tflops": 334.0724812432884,
      "warmup_s": 57.35422158241272
    },
    {
      "mb": 12,
      "ctx": 2048,
      "compile": "default",
      "tok_s": 113402.57779289872,
      "peak_gb": 41.24522496,
      "tflops": 340.9497672701231,
      "warmup_s": 0.641932487487793
    },
    {
      "mb": 16,
      "ctx": 2048,
      "compile": "default",
      "tok_s": 115294.6307953683,
      "peak_gb": 52.63751168,
      "tflops": 346.63830666146606,
      "warmup_s": 0.8390512466430664
    },
    {
      "mb": 24,
      "ctx": 2048,
      "compile": "default",
      "tok_s": 117730.31919758776,
      "peak_gb": 75.25300224,
      "tflops": 353.961309454188,
      "warmup_s": 1.2464354038238525
    },
    {
      "mb": 48,
      "ctx": 1024,
      "compile": "max-autotune",
      "tok_s": 130146.54947516992,
      "peak_gb": 78.26123776,
      "tflops": 391.29124415148357,
      "warmup_s": 161.93337106704712
    }
  ],
  "best": {
    "mb": 48,
    "ctx": 1024,
    "compile": "default",
    "tok_s": 130583.87669116902,
    "peak_gb": 75.252740096,
    "tflops": 392.606087388893,
    "warmup_s": 1.1335768699645996
  }
}