{
  "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 version was not recorded for public reproduction.",
    "Weights file SHA-256 was not recorded for public reproduction."
  ],
  "engine": {
    "name": "llama.cpp CUDA",
    "recipe": "plain",
    "version": null
  },
  "format": "olir/0.1",
  "history": {
    "corrections": [],
    "supersedes": []
  },
  "id": "b911e8d530594210b2fc3d02",
  "measured_at": "2026-10-07T01:06:34.322344+00:00",
  "metrics": [
    {
      "basis": null,
      "n": null,
      "name": "wall_seconds",
      "scope": "machine_total",
      "source_field": "wall_seconds",
      "statistic": "single",
      "unit": "s",
      "value": 629.5073147419607
    },
    {
      "basis": null,
      "n": 5,
      "name": "ttft_median_s",
      "scope": "per_person",
      "source_field": "ttft_median_s",
      "statistic": "median",
      "unit": "s",
      "value": 375.66302481101593
    },
    {
      "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": 624.6775280639995
    },
    {
      "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": 3
    },
    {
      "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": 127.27466071100207
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[0].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 491.53372020000825
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[0].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.24433160324097894
    },
    {
      "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": 1964.38158705999
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[0].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 3
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[0].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250016
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[0].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 90
    },
    {
      "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": 127.26417148596374
    },
    {
      "basis": null,
      "n": null,
      "name": "cached_prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[0].cached_prompt_tokens",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[1].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 249.8476865470293
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[1].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 624.6742157849949
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[1].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.2401110726686633
    },
    {
      "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": 1000.6736642443916
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[1].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 4
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[1].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250016
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[1].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 91
    },
    {
      "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": 249.83982097601984
    },
    {
      "basis": null,
      "n": null,
      "name": "cached_prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[1].cached_prompt_tokens",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[2].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 375.66302481101593
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[2].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 626.8256207300001
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[2].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.3503706420855181
    },
    {
      "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": 665.5326276142696
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[2].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 2
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[2].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250016
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[2].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 89
    },
    {
      "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": 375.6497323960066
    },
    {
      "basis": null,
      "n": null,
      "name": "cached_prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[2].cached_prompt_tokens",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[3].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 501.0691677770228
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[3].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 628.4859969120007
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[3].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.6906466013745953
    },
    {
      "basis": "effective, includes queueing",
      "n": null,
      "name": "prefill_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[3].effective_prefill_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 498.9650453034017
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[3].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 1
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[3].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250016
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[3].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 89
    },
    {
      "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": 501.0680842609727
    },
    {
      "basis": null,
      "n": null,
      "name": "cached_prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[3].cached_prompt_tokens",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "long_context_requests[4].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 624.6775280639995
    },
    {
      "basis": null,
      "n": null,
      "name": "total_s",
      "scope": "per_person",
      "source_field": "long_context_requests[4].total_s",
      "statistic": "single",
      "unit": "s",
      "value": 629.500791908009
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[4].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 18.452235431934398
    },
    {
      "basis": "effective, includes queueing",
      "n": null,
      "name": "prefill_tps",
      "scope": "per_person",
      "source_field": "long_context_requests[4].effective_prefill_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 400.23210179314367
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "long_context_requests[4].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[4].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 250016
    },
    {
      "basis": null,
      "n": null,
      "name": "output_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[4].output_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 90
    },
    {
      "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": 624.672843057022
    },
    {
      "basis": null,
      "n": null,
      "name": "cached_prompt_tokens",
      "scope": "per_person",
      "source_field": "long_context_requests[4].cached_prompt_tokens",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "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": 0.884064732178223
    },
    {
      "basis": null,
      "n": null,
      "name": "aggregate_e2e_tps",
      "scope": "machine_total",
      "source_field": "aggregate_end_to_end_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 0.7132562076487516
    },
    {
      "basis": null,
      "n": null,
      "name": "peak_host_unavailable_gib",
      "scope": "machine_total",
      "source_field": "peak_host_unavailable_gib",
      "statistic": "max",
      "unit": "GiB",
      "value": 36.17306137084961
    }
  ],
  "model": {
    "artifact_sha256": [],
    "publisher_repo": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16",
    "publisher_revision": "a9904d24bcc1d289a1950fa9d2b978c47cf903b9",
    "variant": "Q4_K_M GGUF variant",
    "weights_dtype": "Q4_K_M",
    "weights_format": "GGUF"
  },
  "outcome": {
    "gates": [
      {
        "met": true,
        "name": "requests",
        "observed": "5/5"
      },
      {
        "met": true,
        "name": "hidden-facts",
        "observed": "15/15"
      },
      {
        "met": false,
        "name": "first-token-300-s-reference",
        "observed": 624.6775280639995
      }
    ],
    "reasons": [
      "Observed decode overlap was 3 of 5 requested users."
    ],
    "status": "fail"
  },
  "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": "04cf656671acc024178598a26462c2de38b6f97137ea5abf44a09b83b8047c5e",
    "harness": {
      "name": "keep-long-context-concurrency-v1",
      "version": "1"
    },
    "harness_sha256": null,
    "os": null,
    "workload_name": null,
    "workload_version": null
  },
  "requests": [
    {
      "cached_prompt_tokens": 0,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 0.24433160324097894,
      "effective_prefill_tps": 1964.38158705999,
      "first_real_token_from_dispatch_s": 127.26417148596374,
      "first_token_within_300_seconds": true,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 90,
      "prompt_tokens": 250016,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 491.53372020000825,
      "ttft_s": 127.27466071100207,
      "user_number": 3,
      "valid_json": true
    },
    {
      "cached_prompt_tokens": 0,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 0.2401110726686633,
      "effective_prefill_tps": 1000.6736642443916,
      "first_real_token_from_dispatch_s": 249.83982097601984,
      "first_token_within_300_seconds": true,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 91,
      "prompt_tokens": 250016,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 624.6742157849949,
      "ttft_s": 249.8476865470293,
      "user_number": 4,
      "valid_json": true
    },
    {
      "cached_prompt_tokens": 0,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 0.3503706420855181,
      "effective_prefill_tps": 665.5326276142696,
      "first_real_token_from_dispatch_s": 375.6497323960066,
      "first_token_within_300_seconds": false,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 89,
      "prompt_tokens": 250016,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 626.8256207300001,
      "ttft_s": 375.66302481101593,
      "user_number": 2,
      "valid_json": true
    },
    {
      "cached_prompt_tokens": 0,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 0.6906466013745953,
      "effective_prefill_tps": 498.9650453034017,
      "first_real_token_from_dispatch_s": 501.0680842609727,
      "first_token_within_300_seconds": false,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 89,
      "prompt_tokens": 250016,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 628.4859969120007,
      "ttft_s": 501.0691677770228,
      "user_number": 1,
      "valid_json": true
    },
    {
      "cached_prompt_tokens": 0,
      "checks_passed": true,
      "completed": true,
      "decode_tps": 18.452235431934398,
      "effective_prefill_tps": 400.23210179314367,
      "first_real_token_from_dispatch_s": 624.672843057022,
      "first_token_within_300_seconds": false,
      "needle_correct": 3,
      "needle_total": 3,
      "output_tokens": 90,
      "prompt_tokens": 250016,
      "stop_and_done": true,
      "strict_json": true,
      "total_s": 629.500791908009,
      "ttft_s": 624.6775280639995,
      "user_number": 0,
      "valid_json": true
    }
  ],
  "scenario": {
    "id": "long-context-concurrency",
    "version": "1"
  },
  "settings": {
    "cache_state": "cold",
    "concurrency": 5,
    "configured_context_tokens": 262144,
    "context_tokens": 250000,
    "fixture_variant": "distinct-v1",
    "kv_dtype": "q8_0",
    "max_output_tokens": 512,
    "scenario": "distinct",
    "thinking": false
  },
  "setup_complete": false,
  "setup_id": "9e5118825b2267598d5691416d3ab69e",
  "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": 250016,
      "median": 250016,
      "min": 250016
    },
    "output_tokens": 512,
    "repetitions": null,
    "users": 5,
    "warmup": null
  }
}
