{
  "caveats": [
    "15/16 at both concurrency levels; measured but unqualified. Short prompts and reused caches do not establish long-context quality.",
    "Weights file SHA-256 was not recorded for public reproduction."
  ],
  "engine": {
    "name": "llama.cpp",
    "recipe": "with multi-token prediction",
    "version": "native CUDA 13, sm_121a"
  },
  "format": "olir/0.1",
  "history": {
    "corrections": [],
    "supersedes": []
  },
  "id": "d92bd079da7d9fc5722fc394",
  "measured_at": "2026-10-05T18:22:43.041657+00:00",
  "metrics": [
    {
      "basis": null,
      "n": null,
      "name": "workload_total",
      "scope": "machine_total",
      "source_field": "workload_total",
      "statistic": "single",
      "unit": "count",
      "value": 16
    },
    {
      "basis": null,
      "n": null,
      "name": "chat_decode_tps",
      "scope": "per_person",
      "source_field": "chat_decode_tps",
      "statistic": "mean",
      "unit": "tok/s",
      "value": 26.98740543967368
    },
    {
      "basis": null,
      "n": null,
      "name": "code_ttft_mean_s",
      "scope": "per_person",
      "source_field": "code_ttft_mean_s",
      "statistic": "mean",
      "unit": "s",
      "value": 0.29756888799602166
    },
    {
      "basis": null,
      "n": null,
      "name": "workload_passed",
      "scope": "machine_total",
      "source_field": "workload_passed",
      "statistic": "single",
      "unit": "count",
      "value": 15
    },
    {
      "basis": null,
      "n": null,
      "name": "concurrency",
      "scope": "machine_total",
      "source_field": "workload_levels[0].concurrency",
      "statistic": "single",
      "unit": "count",
      "value": 1
    },
    {
      "basis": null,
      "n": null,
      "name": "passed",
      "scope": "machine_total",
      "source_field": "workload_levels[0].passed",
      "statistic": "single",
      "unit": "count",
      "value": 15
    },
    {
      "basis": null,
      "n": null,
      "name": "total",
      "scope": "machine_total",
      "source_field": "workload_levels[0].total",
      "statistic": "single",
      "unit": "count",
      "value": 16
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_p95_s",
      "scope": "per_person",
      "source_field": "workload_levels[0].ttft_p95_s",
      "statistic": "p95",
      "unit": "s",
      "value": 0.5690766500192694
    },
    {
      "basis": null,
      "n": null,
      "name": "concurrency",
      "scope": "machine_total",
      "source_field": "workload_levels[1].concurrency",
      "statistic": "single",
      "unit": "count",
      "value": 2
    },
    {
      "basis": null,
      "n": null,
      "name": "passed",
      "scope": "machine_total",
      "source_field": "workload_levels[1].passed",
      "statistic": "single",
      "unit": "count",
      "value": 15
    },
    {
      "basis": null,
      "n": null,
      "name": "total",
      "scope": "machine_total",
      "source_field": "workload_levels[1].total",
      "statistic": "single",
      "unit": "count",
      "value": 16
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_p95_s",
      "scope": "per_person",
      "source_field": "workload_levels[1].ttft_p95_s",
      "statistic": "p95",
      "unit": "s",
      "value": 0.794703905005008
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_concurrency",
      "scope": "machine_total",
      "source_field": "aggregate_concurrency",
      "statistic": "single",
      "unit": "count",
      "value": 2
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_e2e_tps",
      "scope": "machine_total",
      "source_field": "aggregate_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 53.585154787958096
    },
    {
      "basis": null,
      "n": null,
      "name": "code_decode_tps",
      "scope": "per_person",
      "source_field": "code_decode_tps",
      "statistic": "mean",
      "unit": "tok/s",
      "value": 33.642655882501664
    },
    {
      "basis": null,
      "n": null,
      "name": "chat_ttft_mean_s",
      "scope": "per_person",
      "source_field": "chat_ttft_mean_s",
      "statistic": "mean",
      "unit": "s",
      "value": 0.2478214856625224
    },
    {
      "basis": null,
      "n": null,
      "name": "peak_host_unavailable_gib",
      "scope": "machine_total",
      "source_field": "peak_host_unavailable_gib",
      "statistic": "max",
      "unit": "GiB",
      "value": 27.523632049560547
    }
  ],
  "model": {
    "artifact_sha256": [],
    "publisher_repo": "Qwen/Qwen3.8-27B",
    "publisher_revision": "1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0",
    "variant": "GSQ-RCO IQ3_S GGUF variant",
    "weights_dtype": "GSQ-RCO IQ3_S",
    "weights_format": "GGUF"
  },
  "outcome": {
    "gates": [
      {
        "met": false,
        "name": "structured-tasks",
        "observed": "15/16"
      }
    ],
    "reasons": [
      "One task returned fenced text instead of one strict JSON value at each concurrency level.",
      "Native runtime identity does not satisfy the workload runner's OCI identity gate."
    ],
    "status": "partial"
  },
  "publication_group": "pre-standard",
  "published": null,
  "publisher": {
    "name": "Skilak Keep"
  },
  "quality": [
    {
      "passed": 15,
      "skipped": null,
      "suite": "structured-tasks",
      "total": 16,
      "version": null
    }
  ],
  "record_date": "2026-10-05",
  "reproduce": {
    "driver": null,
    "fixture_sha256": null,
    "harness": {
      "name": null,
      "version": null
    },
    "harness_sha256": null,
    "os": null,
    "workload_name": null,
    "workload_version": null
  },
  "requests": [],
  "scenario": {
    "id": "workload-qualification",
    "version": null
  },
  "settings": {
    "concurrency": 2,
    "context_tokens": 65536,
    "max_output_tokens": 512
  },
  "setup_complete": false,
  "setup_id": "7b2be9be996767ee320ef3e0abcb3035",
  "standard": "pre-standard",
  "system": {
    "driver": null,
    "hardware": "NVIDIA GB10, 128 GB unified memory",
    "os": null
  },
  "workload": {
    "cache": null,
    "content_tokens": null,
    "document": null,
    "input_tokens": null,
    "output_tokens": 512,
    "repetitions": null,
    "users": 2,
    "warmup": null
  }
}