{
  "caveats": [
    "Lab observation, not customer acceptance. Prompt content length excludes the server chat template; actual per-request usage is recorded. Synthesis is requested and retained but semantic accuracy is unassessed. All requested users must overlap in observed decode to pass."
  ],
  "engine": {
    "name": "NVIDIA NGC vLLM",
    "recipe": "plain",
    "version": "0.29.0+5013de39.dev"
  },
  "format": "olir/0.1",
  "history": {
    "corrections": [],
    "supersedes": []
  },
  "id": "063d452cce7348478144544b",
  "measured_at": "2026-10-06T03:06:31.741169+00:00",
  "metrics": [
    {
      "basis": null,
      "n": null,
      "name": "wall_seconds",
      "scope": "machine_total",
      "source_field": "wall_seconds",
      "statistic": "single",
      "unit": "s",
      "value": 1395.1257285339525
    },
    {
      "basis": null,
      "n": 3,
      "name": "ttft_median_s",
      "scope": "per_person",
      "source_field": "ttft_median_s",
      "statistic": "median",
      "unit": "s",
      "value": 932.0302819130011
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_total",
      "scope": "machine_total",
      "source_field": "needle_total",
      "statistic": "single",
      "unit": "count",
      "value": 9
    },
    {
      "basis": null,
      "n": 3,
      "name": "ttft_max_s",
      "scope": "per_person",
      "source_field": "ttft_max_s",
      "statistic": "max",
      "unit": "s",
      "value": 1379.8438422949985
    },
    {
      "basis": null,
      "n": null,
      "name": "requests_ok",
      "scope": "machine_total",
      "source_field": "requests_ok",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_correct",
      "scope": "machine_total",
      "source_field": "needle_correct",
      "statistic": "single",
      "unit": "count",
      "value": 9
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_overlap",
      "scope": "machine_total",
      "source_field": "decode_overlap",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_concurrency",
      "scope": "machine_total",
      "source_field": "aggregate_concurrency",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[0].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 471.9909305470064
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[0].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 1386.734988627024
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[0].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.1115066002331742
    },
    {
      "basis": "effective, includes queueing",
      "n": null,
      "name": "prefill_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[0].effective_prefill_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 529.6945000834098
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[0].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 2
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[0].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250011
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[0].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 103
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_correct",
      "scope": "per_person",
      "source_field": "long_context_requests[0].needle_correct",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_total",
      "scope": "per_person",
      "source_field": "long_context_requests[0].needle_total",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[1].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 932.0302819130011
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[1].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 1391.490379876981
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[1].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.22199969149006807
    },
    {
      "basis": "effective, includes queueing",
      "n": null,
      "name": "prefill_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[1].effective_prefill_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 268.24343033881905
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[1].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[1].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250011
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[1].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 103
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_correct",
      "scope": "per_person",
      "source_field": "long_context_requests[1].needle_correct",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_total",
      "scope": "per_person",
      "source_field": "long_context_requests[1].needle_total",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[2].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 1379.8438422949985
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[2].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 1395.084906300006
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[2].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 6.692446141981135
    },
    {
      "basis": "effective, includes queueing",
      "n": null,
      "name": "prefill_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[2].effective_prefill_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 181.18789412008684
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[2].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 1
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[2].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250011
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[2].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 103
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_correct",
      "scope": "per_person",
      "source_field": "long_context_requests[2].needle_correct",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_total",
      "scope": "per_person",
      "source_field": "long_context_requests[2].needle_total",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "requests_total",
      "scope": "machine_total",
      "source_field": "requests_total",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_decode_tps",
      "scope": "machine_total",
      "source_field": "aggregate_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.3314937652694489
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_e2e_tps",
      "scope": "machine_total",
      "source_field": "aggregate_end_to_end_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.2214854143108006
    },
    {
      "basis": null,
      "n": null,
      "name": "peak_host_unavailable_gib",
      "scope": "machine_total",
      "source_field": "peak_host_unavailable_gib",
      "statistic": "max",
      "unit": "GiB",
      "value": 103.3937759399414
    }
  ],
  "model": {
    "artifact_sha256": [
      "1d8268aa85ace093a561e3e7b63b9d390dac1cd55a90cd55b5ec509c3c9da9fe",
      "f078f51aed0c1dba613c1759487004ee06aba1a290606b6be125f59ab710959e"
    ],
    "publisher_repo": "Qwen/Qwen3.8-27B",
    "publisher_revision": "1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0",
    "variant": "INT4 variant",
    "weights_dtype": "INT4",
    "weights_format": "safetensors"
  },
  "outcome": {
    "gates": [
      {
        "met": true,
        "name": "requests",
        "observed": "3/3"
      },
      {
        "met": true,
        "name": "hidden-facts",
        "observed": "9/9"
      },
      {
        "met": false,
        "name": "first-token-300-s-reference",
        "observed": 1379.8438422949985
      }
    ],
    "reasons": [],
    "status": "pass"
  },
  "publication_group": "pre-standard",
  "published": null,
  "publisher": {
    "name": "Skilak Keep"
  },
  "quality": [
    {
      "passed": 9,
      "skipped": null,
      "suite": "hidden-facts",
      "total": 9,
      "version": "1"
    }
  ],
  "record_date": "2026-10-06",
  "reproduce": {
    "driver": null,
    "fixture_sha256": "1e01cbcbd88ac9878b889cba44f56c74f5ab023f35e2d28a3bbe343923d909e0",
    "harness": {
      "name": "keep-long-context-concurrency-v1",
      "version": "1"
    },
    "harness_sha256": null,
    "os": null,
    "workload_name": null,
    "workload_version": null
  },
  "requests": [
    {
      "cached_prompt_tokens": null,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 0.1115066002331742,
      "effective_prefill_tps": 529.6945000834098,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 103,
      "prompt_tokens": 250011,
      "total_s": 1386.734988627024,
      "ttft_s": 471.9909305470064,
      "user_number": 2,
      "valid_json": null
    },
    {
      "cached_prompt_tokens": null,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 0.22199969149006807,
      "effective_prefill_tps": 268.24343033881905,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 103,
      "prompt_tokens": 250011,
      "total_s": 1391.490379876981,
      "ttft_s": 932.0302819130011,
      "user_number": 0,
      "valid_json": null
    },
    {
      "cached_prompt_tokens": null,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 6.692446141981135,
      "effective_prefill_tps": 181.18789412008684,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 103,
      "prompt_tokens": 250011,
      "total_s": 1395.084906300006,
      "ttft_s": 1379.8438422949985,
      "user_number": 1,
      "valid_json": null
    }
  ],
  "scenario": {
    "id": "long-context-concurrency",
    "version": "1"
  },
  "settings": {
    "cache_state": "cold",
    "concurrency": 3,
    "configured_context_tokens": 262144,
    "context_tokens": 250000,
    "fixture_variant": "distinct-v1",
    "kv_dtype": "fp8",
    "max_output_tokens": 512,
    "scenario": "distinct",
    "thinking": false
  },
  "setup_complete": false,
  "setup_id": "2a73804cb33351a1318b0917ba061450",
  "standard": "pre-standard",
  "system": {
    "driver": null,
    "hardware": "NVIDIA GB10, 128 GB unified memory",
    "os": null
  },
  "workload": {
    "cache": "cold",
    "content_tokens": 250000,
    "document": "distinct",
    "input_tokens": {
      "max": 250011,
      "median": 250011,
      "min": 250011
    },
    "output_tokens": 512,
    "repetitions": null,
    "users": 3,
    "warmup": null
  }
}