Gemma 4 26B A4B IT benchmark on an M5 · 32 GB

{
  "runId": "run_a5a34310-3767-40c8-8763-c2bdbdbbbb39",
  "bundleId": "mlx-gemma-4-26b-a4b-it-ud-mlx-4bit-6cbe4f",
  "status": "verified",
  "promptTokens": 40960,
  "completionTokens": 10240,
  "contextLength": 5120,
  "harness": {
    "version": "0.1.21",
    "gitSha": "unknown"
  },
  "runtime": {
    "name": "mlx_lm",
    "version": "0.31.2",
    "buildFlags": null
  },
  "model": {
    "displayName": "Gemma 4 26B A4B IT",
    "format": "mlx",
    "quant": "ud-4bit",
    "architecture": "gemma4",
    "source": "unsloth/gemma-4-26b-a4b-it-UD-MLX-4bit",
    "fileSizeBytes": 16104634373,
    "lab": {
      "name": "Google",
      "slug": "google"
    },
    "quantizedBy": {
      "name": "Unsloth",
      "slug": "unsloth"
    }
  },
  "device": {
    "cpu": "Apple M5",
    "cpuCores": 10,
    "gpu": "Apple M5",
    "gpuCores": 10,
    "gpuCount": 1,
    "ramGb": 32,
    "osName": "macOS",
    "osVersion": "26.4.1"
  },
  "decodeTpsMean": 32.3,
  "prefillTpsMean": 1042.9,
  "ttftP50Ms": 3895.82,
  "idleTpsMean": 4448,
  "peakRssMb": 17408,
  "trialsPassed": 10,
  "trialsTotal": 10,
  "runnabilityScore": 0.5498164285714285,
  "bundleSha256": "ab0cef6edf66dd626b8e33fe0cbf2ff05487fcb08a16a0c181c3a2fcea9b1549",
  "createdAt": "2026-04-20T11:17:57.482Z"
}