model.gguf benchmark on an AMD's logo.EPYC

<- Runs

Prompt tokens

8,192

Generation tokens

2,048

Trials passed

2/2

Rejected

300.5 tok/s

1,300.0 tok/s

Peak memory

0.00/4 GB

Runs great

Trials

Decode / Prefill Speeds

Metadata

metadata.json
{
  "runId": "run_48f2fbf4-e59a-4652-9c38-9e2e921a78bb",
  "bundleId": "llamacpp-model.gguf-e32f56",
  "status": "rejected",
  "promptTokens": 8192,
  "completionTokens": 2048,
  "contextLength": 5120,
  "harness": {
    "version": "0.1.12",
    "gitSha": "01f7616"
  },
  "runtime": {
    "name": "llama.cpp",
    "version": "b8240",
    "buildFlags": "metal"
  },
  "model": {
    "displayName": "model.gguf",
    "format": "gguf",
    "quant": null,
    "architecture": null,
    "source": null,
    "fileSizeBytes": 2048,
    "lab": null,
    "quantizedBy": null
  },
  "device": {
    "cpu": "AMD EPYC",
    "cpuCores": 2,
    "gpu": "None",
    "gpuCores": 0,
    "gpuCount": 0,
    "ramGb": 4,
    "osName": "Ubuntu 22.04 LTS",
    "osVersion": "22.04"
  },
  "decodeTpsMean": 300.5,
  "prefillTpsMean": 1300,
  "ttftP50Ms": 3150.77,
  "idleTpsMean": 3,
  "peakRssMb": 3,
  "trialsPassed": 2,
  "trialsTotal": 2,
  "runnabilityScore": 0.881,
  "bundleSha256": "1ff7fca2c9783b057a06b59324e144ba734398b60eaa4514d3ae0cc3888ab36c",
  "createdAt": "2026-09-03T22:28:41.516Z"
}