Qwen3-4B benchmark on a GeForce RTX 4060 Laptop GPU

{
  "runId": "run_47525418-26ed-4d51-bd2d-512d69baeebb",
  "bundleId": "llamacpp-qwen3-4b-q4_k_m.gguf-1607a4",
  "status": "verified",
  "promptTokens": 40960,
  "completionTokens": 10240,
  "contextLength": 5120,
  "harness": {
    "version": "0.1.16",
    "gitSha": "08f2145"
  },
  "runtime": {
    "name": "llama.cpp",
    "version": "b1",
    "buildFlags": "metal"
  },
  "model": {
    "displayName": "Qwen3-4B",
    "format": "gguf",
    "quant": "q4_k_m",
    "architecture": "qwen3",
    "source": "unsloth/Qwen3-4B-GGUF:Qwen3-4B-Q4_K_M.gguf",
    "fileSizeBytes": 2497281312,
    "lab": {
      "name": "Qwen",
      "slug": "qwen"
    },
    "quantizedBy": {
      "name": "Unsloth",
      "slug": "unsloth"
    }
  },
  "device": {
    "cpu": "AMD Ryzen 7 7840HS w/ Radeon 780M Graphics",
    "cpuCores": 16,
    "gpu": "NVIDIA GeForce RTX 4060 Laptop GPU",
    "gpuCores": 1,
    "gpuCount": 1,
    "ramGb": 15,
    "osName": "Arch Linux",
    "osVersion": "Unknown"
  },
  "decodeTpsMean": 64,
  "prefillTpsMean": 1754.9,
  "ttftP50Ms": 2347.52,
  "idleTpsMean": 882.2,
  "peakRssMb": 921.6,
  "trialsPassed": 10,
  "trialsTotal": 10,
  "runnabilityScore": 0.834843,
  "bundleSha256": "39bfc33004364bb721fa3d054ca190fee6028ea7a0f12f4e73d0b8a0624396ab",
  "createdAt": "2026-03-31T00:39:45.136Z"
}