{ "case": { "id": "11-latest-abliterated-80k-mtp2-p010", "model": "abliterated-latest", "context": 80000, "split": [ 72, 28 ], "projector": "gpu", "mtp": true, "mtp_max": 2, "quality": false, "long_fill": 0.0, "vision": true }, "model_path": "/models/qwen3.8-27b-abliterated/Qwen3.8-27B-ABLITERATED-Q4_K_M.gguf", "started": "2026-08-22T21:28:26.289184+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", "-e", "MTMD_BACKEND_DEVICE=CUDA1", "mike-ai/llama.cpp:3f545bec-test", "--model", "/models/qwen3.8-27b-abliterated/Qwen3.8-27B-ABLITERATED-Q4_K_M.gguf", "--alias", "11-latest-abliterated-80k-mtp2-p010", "--ctx-size", "80000", "--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", "72,28", "--mmproj", "/models/qwen3.8-27b-abliterated/mmproj-Qwen3.8-27B-ABLITERATED-F16.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", "--spec-draft-p-min", "0.10" ], "docker_launch_returncode": 0, "docker_launch_stderr": "", "loaded": true, "load_seconds": 8.106, "gpu_after_load": [ { "index": "0", "name": "NVIDIA GeForce RTX 3060", "uuid": "GPU-4834d9d7-5b61-3004-1fb3-4ae49d482d4b", "total_mib": "12288", "used_mib": "7295", "free_mib": "4615", "temperature_c": "50", "power_w": "44.72" }, { "index": "1", "name": "NVIDIA GeForce RTX 5080", "uuid": "GPU-8ad38c6c-5a01-9d8e-1dfa-ed662ad78fbe", "total_mib": "16303", "used_mib": "12524", "free_mib": "3358", "temperature_c": "52", "power_w": "57.75" } ], "speed_runs": [ { "tokens": 384, "timings": { "cache_n": 0, "prompt_n": 28, "prompt_ms": 222.267, "prompt_per_token_ms": 7.938107142857143, "prompt_per_second": 125.97461611485285, "predicted_n": 384, "predicted_ms": 8564.865, "predicted_per_token_ms": 22.362571801566578, "predicted_per_second": 44.717575817015216, "draft_n": 324, "draft_n_accepted": 221 } }, { "tokens": 384, "timings": { "cache_n": 0, "prompt_n": 28, "prompt_ms": 203.501, "prompt_per_token_ms": 7.267892857142857, "prompt_per_second": 137.5914614670198, "predicted_n": 384, "predicted_ms": 8556.474, "predicted_per_token_ms": 22.34066318537859, "predicted_per_second": 44.76142859780793, "draft_n": 324, "draft_n_accepted": 221 } } ], "mean_predicted_tps": 44.739502207411576, "vision": { "ok": true, "elapsed_seconds": 27.45, "content": "", "timings": { "cache_n": 0, "prompt_n": 4358, "prompt_ms": 15624.828, "prompt_per_token_ms": 3.585320789352914, "prompt_per_second": 278.91507029709385, "predicted_n": 500, "predicted_ms": 11572.721, "predicted_per_token_ms": 23.191825651302604, "predicted_per_second": 43.11864081057515, "draft_n": 429, "draft_n_accepted": 284 } }, "status": "complete", "server_log": "/data/benchmarks/qwen38-final-acceptance-20260822/cases/11-latest-abliterated-80k-mtp2-p010/server.log", "oom": false, "finished": "2026-08-22T21:29:20.614813+00:00" }