{
  "case": {
    "id": "09-latest-nvfp4-72k-split-mtp2-p010",
    "model": "nvfp4-latest",
    "context": 72000,
    "split": [
      85,
      15
    ],
    "projector": "off",
    "mtp": true,
    "mtp_max": 2,
    "quality": false,
    "long_fill": 0.0,
    "vision": false
  },
  "model_path": "/models/qwen3.8-nvfp4-split/Qwen3.8-27B-NVFP4-Q4_K_M-mtp.gguf",
  "started": "2026-08-22T21:21:57.882396+00:00",
  "command": [
    "docker",
    "run",
    "-d",
    "--name",
    "mike-ai-qwen38-dualgpu-bench",
    "--gpus",
    "all",
    "--ipc",
    "host",
    "--read-only",
    "--tmpfs",
    "/tmp:size=1g,mode=1777",
    "--security-opt",
    "no-new-privileges:true",
    "--cap-drop",
    "ALL",
    "-e",
    "NVIDIA_VISIBLE_DEVICES=GPU-8ad38c6c-5a01-9d8e-1dfa-ed662ad78fbe,GPU-4834d9d7-5b61-3004-1fb3-4ae49d482d4b",
    "-v",
    "/data/models:/models:ro",
    "-p",
    "127.0.0.1:18099:8080",
    "mike-ai/llama.cpp:3f545bec-test",
    "--model",
    "/models/qwen3.8-nvfp4-split/Qwen3.8-27B-NVFP4-Q4_K_M-mtp.gguf",
    "--alias",
    "09-latest-nvfp4-72k-split-mtp2-p010",
    "--ctx-size",
    "72000",
    "--flash-attn",
    "on",
    "--cache-type-k",
    "q4_0",
    "--cache-type-v",
    "q4_0",
    "--threads",
    "6",
    "--threads-batch",
    "6",
    "--batch-size",
    "64",
    "--ubatch-size",
    "32",
    "--parallel",
    "1",
    "--jinja",
    "--reasoning",
    "auto",
    "--reasoning-budget",
    "768",
    "--host",
    "0.0.0.0",
    "--port",
    "8080",
    "--metrics",
    "--fit",
    "off",
    "--n-gpu-layers",
    "all",
    "--no-mmap",
    "--no-ui",
    "--temperature",
    "0.2",
    "--top-p",
    "0.8",
    "--top-k",
    "20",
    "--device",
    "CUDA0,CUDA1",
    "--main-gpu",
    "0",
    "--split-mode",
    "layer",
    "--tensor-split",
    "85,15",
    "--spec-type",
    "draft-mtp",
    "--spec-draft-n-max",
    "2",
    "--spec-draft-type-k",
    "f16",
    "--spec-draft-type-v",
    "f16",
    "--spec-draft-p-min",
    "0.10",
    "--model-draft",
    "/models/qwen3.8-nvfp4-split/mtp-Qwen3.8-27B-NVFP4.gguf",
    "--device-draft",
    "CUDA1",
    "--n-gpu-layers-draft",
    "999"
  ],
  "docker_launch_returncode": 0,
  "docker_launch_stderr": "",
  "loaded": true,
  "load_seconds": 6.523,
  "gpu_after_load": [
    {
      "index": "0",
      "name": "NVIDIA GeForce RTX 3060",
      "uuid": "GPU-4834d9d7-5b61-3004-1fb3-4ae49d482d4b",
      "total_mib": "12288",
      "used_mib": "6909",
      "free_mib": "5001",
      "temperature_c": "46",
      "power_w": "43.49"
    },
    {
      "index": "1",
      "name": "NVIDIA GeForce RTX 5080",
      "uuid": "GPU-8ad38c6c-5a01-9d8e-1dfa-ed662ad78fbe",
      "total_mib": "16303",
      "used_mib": "13812",
      "free_mib": "2070",
      "temperature_c": "48",
      "power_w": "54.92"
    }
  ],
  "speed_runs": [
    {
      "tokens": 384,
      "timings": {
        "cache_n": 0,
        "prompt_n": 28,
        "prompt_ms": 258.302,
        "prompt_per_token_ms": 9.225071428571429,
        "prompt_per_second": 108.40024467483798,
        "predicted_n": 384,
        "predicted_ms": 9070.236,
        "predicted_per_token_ms": 23.682078328981724,
        "predicted_per_second": 42.22602366685938,
        "draft_n": 326,
        "draft_n_accepted": 220
      }
    },
    {
      "tokens": 384,
      "timings": {
        "cache_n": 0,
        "prompt_n": 28,
        "prompt_ms": 121.442,
        "prompt_per_token_ms": 4.3372142857142855,
        "prompt_per_second": 230.56273776782336,
        "predicted_n": 384,
        "predicted_ms": 9048.942,
        "predicted_per_token_ms": 23.626480417754568,
        "predicted_per_second": 42.32539008427726,
        "draft_n": 326,
        "draft_n_accepted": 220
      }
    }
  ],
  "mean_predicted_tps": 42.27570687556832,
  "status": "complete",
  "server_log": "/data/benchmarks/qwen38-final-acceptance-20260822/cases/09-latest-nvfp4-72k-split-mtp2-p010/server.log",
  "oom": false,
  "finished": "2026-08-22T21:22:24.496624+00:00"
}