Qwen3.6-27B benchmark on an M1 Max · 32 GB

{
  "runId": "run_780da56e-2786-4d69-b050-31ea3b91b9d9",
  "bundleId": "llamacpp-qwen3.6-27b-q4_k_m.gguf-b668b9",
  "status": "verified",
  "promptTokens": 40960,
  "completionTokens": 10240,
  "contextLength": 5120,
  "harness": {
    "version": "0.1.21",
    "gitSha": "unknown"
  },
  "runtime": {
    "name": "llama.cpp",
    "version": "b8180",
    "buildFlags": "metal"
  },
  "model": {
    "displayName": "Qwen3.6-27B",
    "format": "gguf",
    "quant": "q4_k_m",
    "architecture": "qwen35",
    "source": "unsloth/Qwen3.6-27B-GGUF:Qwen3.6-27B-Q4_K_M.gguf",
    "fileSizeBytes": 16817244384,
    "lab": {
      "name": "Qwen",
      "slug": "qwen"
    },
    "quantizedBy": {
      "name": "Unsloth",
      "slug": "unsloth"
    }
  },
  "device": {
    "cpu": "Apple M1 Max",
    "cpuCores": 10,
    "gpu": "Apple M1 Max",
    "gpuCores": 32,
    "gpuCount": 1,
    "ramGb": 32,
    "osName": "macOS",
    "osVersion": "26.3"
  },
  "decodeTpsMean": 8.3,
  "prefillTpsMean": 117.8,
  "ttftP50Ms": 34739.24,
  "idleTpsMean": 3771.5,
  "peakRssMb": 4160,
  "trialsPassed": 10,
  "trialsTotal": 10,
  "runnabilityScore": 0.4244415178571429,
  "bundleSha256": "08c78e7701447c323c5f68e8ab80a5536ac944a7d3af5a23ce879df389c81808",
  "createdAt": "2026-05-13T16:45:26.643Z"
}