{
  "prompt_file": "prompt.txt",
  "prompt_sha256": "bc908fefb54858b62757f3193b79bd1e812380da5a9dbb5598e62b3e1522fedc",
  "request": {
    "temperature": 0.7,
    "top_p": 0.95,
    "thinking": "enabled",
    "system_message": false,
    "tools": false,
    "stream": true,
    "max_tokens": null,
    "request_timeout_seconds": null
  },
  "matched_conditions": [
    "exact same prompt bytes",
    "one direct request per model",
    "no output-token cap",
    "no request wall timeout",
    "original outputs preserved",
    "display repairs separate and disclosed"
  ],
  "flash": {
    "model": "GLM-5.3 Flash",
    "weights_quant": "NVFP4",
    "hardware": "2x DGX Spark",
    "tensor_parallel": 2,
    "speculative_decoding": "DFlash2 K7",
    "context_tokens": 262144
  },
  "full": {
    "model": "GLM-5.3 Full 743B",
    "weights_quant": "Int4/Int8Mix",
    "hardware": "4x DGX Spark",
    "tensor_parallel": 4,
    "speculative_decoding": "in-checkpoint MTP K5",
    "kv_cache": "NVFP4 DSA MLA",
    "context_tokens": 300000,
    "checkpoint_repository": "Tech2wild/GLM-5.3-Int4-Int8Mix",
    "checkpoint_revision": "206507bbb047d8223964a0414cd83230c59428f9",
    "weight_shards": 282,
    "weight_bytes": 405241870672,
    "published_sha256_verified": true,
    "quant_structure_verified": true
  }
}