{
  "case": {
    "id": "04-latest-fast-76k-mtp2-p000",
    "model": "iq4-mix",
    "context": 76800,
    "split": null,
    "projector": "gpu",
    "mtp": true,
    "mtp_max": 2,
    "quality": false,
    "long_fill": 0.0,
    "vision": false
  },
  "model_path": "/models/qwen3.8-27b-iq4-mix/Qwen3.8-27B-IQ4-MIX.gguf",
  "started": "2026-08-22T21:20:55.266458+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",
    "-v",
    "/data/models:/models:ro",
    "-p",
    "127.0.0.1:18099:8080",
    "-e",
    "MTMD_BACKEND_DEVICE=CUDA1",
    "mike-ai/llama.cpp:3f545bec-test",
    "--model",
    "/models/qwen3.8-27b-iq4-mix/Qwen3.8-27B-IQ4-MIX.gguf",
    "--alias",
    "04-latest-fast-76k-mtp2-p000",
    "--ctx-size",
    "76800",
    "--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",
    "--split-mode",
    "none",
    "--mmproj",
    "/models/qwen3.8-27b-nvfp4/mmproj-BF16.gguf",
    "--image-min-tokens",
    "1024",
    "--spec-type",
    "draft-mtp",
    "--spec-draft-n-max",
    "2",
    "--spec-draft-type-k",
    "f16",
    "--spec-draft-type-v",
    "f16",
    "--mmproj-device",
    "CUDA1"
  ],
  "docker_launch_returncode": 0,
  "docker_launch_stderr": "",
  "loaded": true,
  "load_seconds": 4.056,
  "gpu_after_load": [
    {
      "index": "0",
      "name": "NVIDIA GeForce RTX 3060",
      "uuid": "GPU-4834d9d7-5b61-3004-1fb3-4ae49d482d4b",
      "total_mib": "12288",
      "used_mib": "1251",
      "free_mib": "10659",
      "temperature_c": "42",
      "power_w": "37.55"
    },
    {
      "index": "1",
      "name": "NVIDIA GeForce RTX 5080",
      "uuid": "GPU-8ad38c6c-5a01-9d8e-1dfa-ed662ad78fbe",
      "total_mib": "16303",
      "used_mib": "15796",
      "free_mib": "86",
      "temperature_c": "55",
      "power_w": "63.49"
    }
  ],
  "speed_runs": [
    {
      "tokens": 384,
      "timings": {
        "cache_n": 0,
        "prompt_n": 28,
        "prompt_ms": 153.99,
        "prompt_per_token_ms": 5.499642857142858,
        "prompt_per_second": 181.82998896032208,
        "predicted_n": 384,
        "predicted_ms": 4477.73,
        "predicted_per_token_ms": 11.691201044386421,
        "predicted_per_second": 85.53441140935251,
        "draft_n": 348,
        "draft_n_accepted": 209
      }
    },
    {
      "tokens": 384,
      "timings": {
        "cache_n": 0,
        "prompt_n": 28,
        "prompt_ms": 101.423,
        "prompt_per_token_ms": 3.62225,
        "prompt_per_second": 276.0715025191525,
        "predicted_n": 384,
        "predicted_ms": 4466.939,
        "predicted_per_token_ms": 11.663026109660576,
        "predicted_per_second": 85.7410410126487,
        "draft_n": 348,
        "draft_n_accepted": 209
      }
    }
  ],
  "mean_predicted_tps": 85.6377262110006,
  "status": "complete",
  "server_log": "/data/benchmarks/qwen38-final-acceptance-20260822/cases/04-latest-fast-76k-mtp2-p000/server.log",
  "oom": false,
  "finished": "2026-08-22T21:21:09.682589+00:00"
}