{
  "key": "mxfp4",
  "mode": "mxfp4",
  "bits": 4,
  "group_size": 32,
  "scope": "Complete artifact and native dequantized-value checks, plus eight authored cached greedy fixtures; not a broad benchmark.",
  "backend": "Linux native MLX CUDA 12 on NVIDIA B200",
  "mac_acceptance": "pending",
  "weight_bytes": 18426513876,
  "files": [
    {
      "file": "model-00001-of-00004.safetensors",
      "bytes": 5349133873,
      "sha256": "81609e9859214b4699e7123c9aca76a6635d1f0f401269a593e6458bf56ea5e6"
    },
    {
      "file": "model-00002-of-00004.safetensors",
      "bytes": 5365849599,
      "sha256": "9fb7b67e0dc5f00cf6f10346f26f61a4bd539931fb0415bcb4e0c84af24d8d43"
    },
    {
      "file": "model-00003-of-00004.safetensors",
      "bytes": 5365849547,
      "sha256": "d7e14ad72cda67145d4103e23c6e573acaa11a94d0891c2197d12bb17df1f1e1"
    },
    {
      "file": "model-00004-of-00004.safetensors",
      "bytes": 2345680857,
      "sha256": "f997e3273452a19aa6a1c12c48a4fd66ff5103292733e70a7f899c182e00bab5"
    }
  ],
  "strict_stock_reload": true,
  "all_stored_floating_values_finite": true,
  "all_dequantized_values_finite": true,
  "all_generation_logits_finite": true,
  "functional_passed": 8,
  "functional_total": 8,
  "tokenizer_source_files": {
    "tokenizer.json": "06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523",
    "tokenizer_config.json": "792fa3f0cb88b111e54ef3134c873531008c4df471d108da17903426e308aa7b",
    "chat_template.jinja": "e84f32a23fdda27689f868aa4a1a5621f41133e51a48d7f3efcbea2839574259"
  },
  "quantized_modules": 512,
  "dequantized_rows_checked": 32631720,
  "http_smoke": {
    "passed": 2,
    "total": 2,
    "endpoint": "/v1/chat/completions"
  },
  "held_out_perplexity": 8.888629727327855,
  "paired_bf16_perplexity": 8.307409052128223,
  "held_out_ppl_ratio_to_bf16": 1.0699641334082053
}