{
  "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": "with multi-token prediction",
    "version": "0.29.0+5013de39.dev"
  },
  "format": "olir/0.1",
  "history": {
    "corrections": [],
    "supersedes": []
  },
  "id": "d0edaadb8a1940038d9fd76e",
  "measured_at": "2026-10-07T07:22:01.496596+00:00",
  "metrics": [
    {
      "basis": null,
      "n": null,
      "name": "wall_seconds",
      "scope": "machine_total",
      "source_field": "wall_seconds",
      "statistic": "single",
      "unit": "s",
      "value": 18.157898496021517
    },
    {
      "basis": null,
      "n": 5,
      "name": "ttft_median_s",
      "scope": "per_person",
      "source_field": "ttft_median_s",
      "statistic": "median",
      "unit": "s",
      "value": 14.437703111965675
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_total",
      "scope": "machine_total",
      "source_field": "needle_total",
      "statistic": "single",
      "unit": "count",
      "value": 15
    },
    {
      "basis": null,
      "n": 5,
      "name": "ttft_max_s",
      "scope": "per_person",
      "source_field": "ttft_max_s",
      "statistic": "max",
      "unit": "s",
      "value": 14.450682270049583
    },
    {
      "basis": null,
      "n": null,
      "name": "requests_ok",
      "scope": "machine_total",
      "source_field": "requests_ok",
      "statistic": "single",
      "unit": "count",
      "value": 5
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_correct",
      "scope": "machine_total",
      "source_field": "needle_correct",
      "statistic": "single",
      "unit": "count",
      "value": 15
    },
    {
      "basis": null,
      "n": null,
      "name": "latency_bar_seconds",
      "scope": "machine_total",
      "source_field": "latency_bar_seconds",
      "statistic": "single",
      "unit": "s",
      "value": 300
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_overlap",
      "scope": "machine_total",
      "source_field": "decode_overlap",
      "statistic": "single",
      "unit": "count",
      "value": 5
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_concurrency",
      "scope": "machine_total",
      "source_field": "aggregate_concurrency",
      "statistic": "single",
      "unit": "count",
      "value": 5
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[0].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 14.450402253016364
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[0].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 17.44542896101484
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[0].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 36.05977860283406
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[0].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 1
    },
    {
      "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": 109
    },
    {
      "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": "first_real_token_from_dispatch_s",
      "scope": "per_person",
      "source_field": "long_context_requests[0].first_real_token_from_dispatch_s",
      "statistic": "single",
      "unit": "s",
      "value": 14.429424229019787
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[1].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 10.89551008300623
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[1].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 17.683375458000228
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[1].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 17.973249800952022
    },
    {
      "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": 123
    },
    {
      "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": "first_real_token_from_dispatch_s",
      "scope": "per_person",
      "source_field": "long_context_requests[1].first_real_token_from_dispatch_s",
      "statistic": "single",
      "unit": "s",
      "value": 10.892245573049877
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[2].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 13.266313811996952
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[2].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 17.72674516902771
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[2].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 27.127421165056983
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[2].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "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": 122
    },
    {
      "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": "first_real_token_from_dispatch_s",
      "scope": "per_person",
      "source_field": "long_context_requests[2].first_real_token_from_dispatch_s",
      "statistic": "single",
      "unit": "s",
      "value": 13.259319078992121
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[3].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 14.450682270049583
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[3].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 17.95352709305007
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[3].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 37.11269170315226
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[3].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 4
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[3].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250011
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[3].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 131
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_correct",
      "scope": "per_person",
      "source_field": "long_context_requests[3].needle_correct",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_total",
      "scope": "per_person",
      "source_field": "long_context_requests[3].needle_total",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "first_real_token_from_dispatch_s",
      "scope": "per_person",
      "source_field": "long_context_requests[3].first_real_token_from_dispatch_s",
      "statistic": "single",
      "unit": "s",
      "value": 14.437643639044836
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[4].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 14.437703111965675
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[4].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 18.135093041986693
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[4].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 37.323626290942904
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[4].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 2
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[4].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250011
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[4].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 139
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_correct",
      "scope": "per_person",
      "source_field": "long_context_requests[4].needle_correct",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "needle_total",
      "scope": "per_person",
      "source_field": "long_context_requests[4].needle_total",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "first_real_token_from_dispatch_s",
      "scope": "per_person",
      "source_field": "long_context_requests[4].first_real_token_from_dispatch_s",
      "statistic": "single",
      "unit": "s",
      "value": 14.432861714973114
    },
    {
      "basis": null,
      "n": null,
      "name": "requests_total",
      "scope": "machine_total",
      "source_field": "requests_total",
      "statistic": "single",
      "unit": "count",
      "value": 5
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_decode_tps",
      "scope": "machine_total",
      "source_field": "aggregate_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 85.36082827820009
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_e2e_tps",
      "scope": "machine_total",
      "source_field": "aggregate_end_to_end_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 34.365210276768615
    },
    {
      "basis": null,
      "n": null,
      "name": "peak_host_unavailable_gib",
      "scope": "machine_total",
      "source_field": "peak_host_unavailable_gib",
      "statistic": "max",
      "unit": "GiB",
      "value": 94.23422622680664
    }
  ],
  "model": {
    "artifact_sha256": [
      "648fe83171485f4be03ea58d86089e088c94c2a3dc88922233f9002f30d8bc8b",
      "8303cef7d375021d69dcafc60d3fad5f8186995b18b0db5e2a609aadeb8a4f51",
      "cb5c5978721a1e87406ae68f1cd5a855326e40ff8bc13832644c392115336987"
    ],
    "publisher_repo": "ornith-ai/Ornith-1.5-35B-A3B",
    "publisher_revision": "10fbf86fed7ecee4a061f8b499a618f46001cac1",
    "variant": "NVFP4 variant",
    "weights_dtype": "NVFP4",
    "weights_format": "safetensors"
  },
  "outcome": {
    "gates": [
      {
        "met": true,
        "name": "requests",
        "observed": "5/5"
      },
      {
        "met": true,
        "name": "hidden-facts",
        "observed": "15/15"
      },
      {
        "met": true,
        "name": "first-token-300-s-reference",
        "observed": 14.450682270049583
      }
    ],
    "reasons": [],
    "status": "pass"
  },
  "publication_group": "pre-standard",
  "published": null,
  "publisher": {
    "name": "Skilak Keep"
  },
  "quality": [
    {
      "passed": 15,
      "skipped": null,
      "suite": "hidden-facts",
      "total": 15,
      "version": "1"
    }
  ],
  "record_date": "2026-10-07",
  "reproduce": {
    "driver": null,
    "fixture_sha256": "77fda79ee1961c35a8f8a77e7d5a01d5f9ce414975e00e640a166e71e5f9093e",
    "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": 36.05977860283406,
      "effective_prefill_tps": null,
      "first_real_token_from_dispatch_s": 14.429424229019787,
      "first_token_within_300_seconds": true,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 109,
      "prompt_tokens": 250011,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 17.44542896101484,
      "ttft_s": 14.450402253016364,
      "user_number": 1,
      "valid_json": true
    },
    {
      "cached_prompt_tokens": null,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 17.973249800952022,
      "effective_prefill_tps": null,
      "first_real_token_from_dispatch_s": 10.892245573049877,
      "first_token_within_300_seconds": true,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 123,
      "prompt_tokens": 250011,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 17.683375458000228,
      "ttft_s": 10.89551008300623,
      "user_number": 0,
      "valid_json": true
    },
    {
      "cached_prompt_tokens": null,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 27.127421165056983,
      "effective_prefill_tps": null,
      "first_real_token_from_dispatch_s": 13.259319078992121,
      "first_token_within_300_seconds": true,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 122,
      "prompt_tokens": 250011,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 17.72674516902771,
      "ttft_s": 13.266313811996952,
      "user_number": 3,
      "valid_json": true
    },
    {
      "cached_prompt_tokens": null,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 37.11269170315226,
      "effective_prefill_tps": null,
      "first_real_token_from_dispatch_s": 14.437643639044836,
      "first_token_within_300_seconds": true,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 131,
      "prompt_tokens": 250011,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 17.95352709305007,
      "ttft_s": 14.450682270049583,
      "user_number": 4,
      "valid_json": true
    },
    {
      "cached_prompt_tokens": null,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 37.323626290942904,
      "effective_prefill_tps": null,
      "first_real_token_from_dispatch_s": 14.432861714973114,
      "first_token_within_300_seconds": true,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 139,
      "prompt_tokens": 250011,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 18.135093041986693,
      "ttft_s": 14.437703111965675,
      "user_number": 2,
      "valid_json": true
    }
  ],
  "scenario": {
    "id": "long-context-concurrency",
    "version": "1"
  },
  "settings": {
    "cache_state": "warm",
    "concurrency": 5,
    "configured_context_tokens": 262144,
    "context_tokens": 250000,
    "fixture_variant": "shared-record-groups-v2",
    "kv_dtype": "fp8",
    "max_output_tokens": 512,
    "scenario": "shared",
    "speculative_decoding": "with multi-token prediction",
    "thinking": false
  },
  "setup_complete": false,
  "setup_id": "bde5d43ff70430becda39144f69c34e0",
  "standard": "pre-standard",
  "system": {
    "driver": null,
    "hardware": "NVIDIA GB10, 128 GB unified memory",
    "os": null
  },
  "workload": {
    "cache": "warm",
    "content_tokens": 250000,
    "document": "shared",
    "input_tokens": {
      "max": 250011,
      "median": 250011,
      "min": 250011
    },
    "output_tokens": 512,
    "repetitions": null,
    "users": 5,
    "warmup": null
  }
}
