LFM2-350M benchmark on an M1 Max · 64 GB

{
  "runId": "run_4db3f4c0-b618-4ea1-8c8f-8632b958f439",
  "bundleId": "llamacpp-lfm2-350m-q4_k_m.gguf-66a9ab",
  "status": "verified",
  "promptTokens": 40960,
  "completionTokens": 10240,
  "contextLength": 5120,
  "harness": {
    "version": "0.1.21",
    "gitSha": "753787f"
  },
  "runtime": {
    "name": "llama.cpp",
    "version": "b8640",
    "buildFlags": "metal"
  },
  "model": {
    "displayName": "LFM2-350M",
    "format": "gguf",
    "quant": "q4_k_m",
    "architecture": "lfm2",
    "source": "LiquidAI/LFM2-350M-GGUF:LFM2-350M-Q4_K_M.gguf",
    "fileSizeBytes": 229309376,
    "lab": {
      "name": "Liquid AI",
      "slug": "liquid"
    },
    "quantizedBy": {
      "name": "Liquid AI",
      "slug": "liquid"
    }
  },
  "device": {
    "cpu": "Apple M1 Max",
    "cpuCores": 10,
    "gpu": "Apple M1 Max",
    "gpuCores": 32,
    "gpuCount": 1,
    "ramGb": 64,
    "osName": "macOS",
    "osVersion": "15.6.1"
  },
  "decodeTpsMean": 400.5,
  "prefillTpsMean": 9083,
  "ttftP50Ms": 441.1,
  "idleTpsMean": 727.3,
  "peakRssMb": 775.9,
  "trialsPassed": 10,
  "trialsTotal": 10,
  "runnabilityScore": 0.994926016671317,
  "bundleSha256": "7158c7377935272093480ba3780fc2bffcff1c6679d72433f4917c9f7e4243ff",
  "createdAt": "2026-04-14T16:28:09.992Z"
}