Qwen3.6-35B-A3B benchmark on an M5 Pro · 64 GB

{
  "runId": "run_b3d85841-4c1f-4d05-b0e6-48872e150cc4",
  "bundleId": "llamacpp-qwen36_35b_iq4_xs.gguf-6ed648",
  "status": "verified",
  "promptTokens": 40960,
  "completionTokens": 10240,
  "contextLength": 5120,
  "harness": {
    "version": "0.1.22",
    "gitSha": "50c24e0"
  },
  "runtime": {
    "name": "llama.cpp",
    "version": "b8920",
    "buildFlags": "metal"
  },
  "model": {
    "displayName": "Qwen3.6-35B-A3B",
    "format": "gguf",
    "quant": "iq4_xs",
    "architecture": "qwen35moe",
    "source": "DuoNeural/Qwen3.6-35B-A3B-Code-imatrix-GGUF:qwen36_35b_IQ4_XS.gguf",
    "fileSizeBytes": 18728777984,
    "lab": {
      "name": "Qwen",
      "slug": "qwen"
    },
    "quantizedBy": {
      "name": "DuoNeural",
      "slug": "duoneural"
    }
  },
  "device": {
    "cpu": "Apple M5 Pro",
    "cpuCores": 18,
    "gpu": "Apple M5 Pro",
    "gpuCores": 20,
    "gpuCount": 1,
    "ramGb": 64,
    "osName": "macOS",
    "osVersion": "26.5.1"
  },
  "decodeTpsMean": 73.5,
  "prefillTpsMean": 1504,
  "ttftP50Ms": 2724.33,
  "idleTpsMean": 9118.4,
  "peakRssMb": 9125.2,
  "trialsPassed": 10,
  "trialsTotal": 10,
  "runnabilityScore": 0.825775927734375,
  "bundleSha256": "7486bbdabab83ae695c98abc790fe7692e37c14a34b3939a2add740f20ff3aeb",
  "createdAt": "2026-06-15T22:04:10.219Z"
}