{
  "base_url": "redacted://routed-capability",
  "clock_domain_id": null,
  "completed": 20,
  "concurrency": 1,
  "configuration_fingerprint": null,
  "context_distribution": {
    "8192": 20
  },
  "context_seed": 941311,
  "context_tokens": 8192,
  "controlled_output_policy": "observe",
  "engine": "vllm-exl3-v84",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-10-17T00:51:49Z",
  "finished_at_unix_ns": 1792198309212628493,
  "gpu": "dual-rtx-pro-6000-maxq",
  "manifest": null,
  "max_context_tokens": null,
  "max_tokens": 16384,
  "measurement_protocol": "capacity-v4-reasoning",
  "metric_population": {
    "aggregate_throughput_eligible": true,
    "attempted_requests": 20,
    "excluded_requests": 0,
    "failure_policy": "failed or contract-invalid responses retain sanitized timing and metadata but are excluded from performance distributions",
    "p99_interpretation": "descriptive-nearest-rank-only",
    "p99_interpretation_by_metric": {
      "decode_tok_s": "descriptive-nearest-rank-only",
      "e2e": "descriptive-nearest-rank-only",
      "effective_prefill_tok_s": "descriptive-nearest-rank-only",
      "tpot": "descriptive-nearest-rank-only",
      "ttft": "descriptive-nearest-rank-only"
    },
    "p99_minimum_recommended_samples": 100,
    "p99_samples_by_metric": {
      "decode_tok_s": 20,
      "e2e": 20,
      "effective_prefill_tok_s": 20,
      "tpot": 20,
      "ttft": 20
    },
    "performance_eligible_requests": 20,
    "successful_requests": 20
  },
  "metrics": {
    "completion_tokens_max": 455.0,
    "completion_tokens_mean": 230.4,
    "completion_tokens_mean_ci95_high": 268.4122998819565,
    "completion_tokens_mean_ci95_low": 192.3877001180435,
    "completion_tokens_min": 114.0,
    "completion_tokens_p25": 175.0,
    "completion_tokens_p50": 209.0,
    "completion_tokens_p75": 260.0,
    "completion_tokens_p90": 330.0,
    "completion_tokens_p95": 395.0,
    "completion_tokens_p99": 455.0,
    "completion_tokens_samples": 20,
    "completion_tokens_stddev": 86.73274134404281,
    "content_chunks": 984,
    "content_chunks_s": null,
    "controlled_output_exact_adherence": 0,
    "controlled_output_exact_adherence_rate": null,
    "controlled_output_samples": 0,
    "controlled_output_unobservable": 0,
    "decode_tok_s_max": 77.07028222679155,
    "decode_tok_s_mean": 70.6972559719683,
    "decode_tok_s_mean_ci95_high": 73.48366213287295,
    "decode_tok_s_mean_ci95_low": 67.91084981106366,
    "decode_tok_s_min": 50.957522003043046,
    "decode_tok_s_p25": 67.98379067005119,
    "decode_tok_s_p50": 68.22103878874331,
    "decode_tok_s_p75": 76.70721268187934,
    "decode_tok_s_p90": 76.96356158280922,
    "decode_tok_s_p95": 76.96680791379138,
    "decode_tok_s_p99": 77.07028222679155,
    "decode_tok_s_samples": 20,
    "decode_tok_s_stddev": 6.357748559905121,
    "e2e_max_ms": 7965.677306987345,
    "e2e_mean_ci95_high_ms": 5138.1103944981305,
    "e2e_mean_ci95_low_ms": 4011.4627499698827,
    "e2e_mean_ms": 4574.786572234007,
    "e2e_min_ms": 2812.4596489360556,
    "e2e_p25_ms": 3573.050903971307,
    "e2e_p50_ms": 4312.904532067478,
    "e2e_p75_ms": 4745.22425502073,
    "e2e_p90_ms": 6152.306975913234,
    "e2e_p95_ms": 7113.6115049012005,
    "e2e_p99_ms": 7965.677306987345,
    "e2e_samples": 20,
    "e2e_stddev_ms": 1285.3371019669298,
    "effective_prefill_tok_s_max": 4736.022505195007,
    "effective_prefill_tok_s_mean": 4639.152154222617,
    "effective_prefill_tok_s_mean_ci95_high": 4667.729574689621,
    "effective_prefill_tok_s_mean_ci95_low": 4610.574733755614,
    "effective_prefill_tok_s_min": 4560.835861679433,
    "effective_prefill_tok_s_p25": 4571.18359913417,
    "effective_prefill_tok_s_p50": 4614.319839754458,
    "effective_prefill_tok_s_p75": 4694.129932684088,
    "effective_prefill_tok_s_p90": 4714.898025320386,
    "effective_prefill_tok_s_p95": 4733.350690424071,
    "effective_prefill_tok_s_p99": 4736.022505195007,
    "effective_prefill_tok_s_samples": 20,
    "effective_prefill_tok_s_stddev": 65.20515794470649,
    "generation_max_ms": 6643.010503030382,
    "generation_mean_ci95_high_ms": 3824.8341562305645,
    "generation_mean_ci95_low_ms": 2705.7968396730653,
    "generation_mean_ms": 3265.315497951815,
    "generation_min_ms": 1530.4925189120695,
    "generation_p25_ms": 2275.22538905032,
    "generation_p50_ms": 2995.267217978835,
    "generation_p75_ms": 3430.428899009712,
    "generation_p90_ms": 4828.3210339723155,
    "generation_p95_ms": 5782.321467995644,
    "generation_p99_ms": 6643.010503030382,
    "generation_samples": 20,
    "generation_stddev_ms": 1276.6548516232237,
    "mean_inter_token_latency_ms_max": 19.624188160881975,
    "mean_inter_token_latency_ms_mean": 14.271905128181833,
    "mean_inter_token_latency_ms_mean_ci95_high": 14.93321504427994,
    "mean_inter_token_latency_ms_mean_ci95_low": 13.610595212083727,
    "mean_inter_token_latency_ms_min": 12.975169820415879,
    "mean_inter_token_latency_ms_p25": 13.018094569219855,
    "mean_inter_token_latency_ms_p50": 14.652412307138244,
    "mean_inter_token_latency_ms_p75": 14.694825751229939,
    "mean_inter_token_latency_ms_p90": 14.755010926004115,
    "mean_inter_token_latency_ms_p95": 14.824924584626492,
    "mean_inter_token_latency_ms_p99": 19.624188160881975,
    "mean_inter_token_latency_ms_samples": 20,
    "mean_inter_token_latency_ms_stddev": 1.508912169989848,
    "mean_time_per_output_token_max_ms": 19.624188160881975,
    "mean_time_per_output_token_mean_ci95_high_ms": 14.93321504427994,
    "mean_time_per_output_token_mean_ci95_low_ms": 13.610595212083727,
    "mean_time_per_output_token_mean_ms": 14.271905128181833,
    "mean_time_per_output_token_min_ms": 12.975169820415879,
    "mean_time_per_output_token_p25_ms": 13.018094569219855,
    "mean_time_per_output_token_p50_ms": 14.652412307138244,
    "mean_time_per_output_token_p75_ms": 14.694825751229939,
    "mean_time_per_output_token_p90_ms": 14.755010926004115,
    "mean_time_per_output_token_p95_ms": 14.824924584626492,
    "mean_time_per_output_token_p99_ms": 19.624188160881975,
    "mean_time_per_output_token_samples": 20,
    "mean_time_per_output_token_stddev_ms": 1.508912169989848,
    "output_token_sources": {
      "usage": 20
    },
    "output_tokens": 4608,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 20,
    "prompt_tokens": 121473,
    "prompt_tokens_max": 6080.0,
    "prompt_tokens_mean": 6073.65,
    "prompt_tokens_mean_ci95_high": 6075.500016842028,
    "prompt_tokens_mean_ci95_low": 6071.799983157971,
    "prompt_tokens_min": 6067.0,
    "prompt_tokens_p25": 6070.0,
    "prompt_tokens_p50": 6073.0,
    "prompt_tokens_p75": 6077.0,
    "prompt_tokens_p90": 6079.0,
    "prompt_tokens_p95": 6079.0,
    "prompt_tokens_p99": 6080.0,
    "prompt_tokens_samples": 20,
    "prompt_tokens_stddev": 4.221187161525958,
    "reasoning_chunks": 3473,
    "request_canary_passed": 0,
    "request_canary_samples": 0,
    "throughput_performance_eligible": true,
    "throughput_tok_s": 50.35569303007099,
    "time_to_first_output_max_ms": 1332.4311999604106,
    "time_to_first_output_mean_ci95_high_ms": 1317.8617740033565,
    "time_to_first_output_mean_ci95_low_ms": 1301.080374561027,
    "time_to_first_output_mean_ms": 1309.4710742821917,
    "time_to_first_output_min_ms": 1281.6662068944424,
    "time_to_first_output_p25_ms": 1289.327758946456,
    "time_to_first_output_p50_ms": 1314.7953560110182,
    "time_to_first_output_p75_ms": 1327.839085017331,
    "time_to_first_output_p90_ms": 1331.290036905557,
    "time_to_first_output_p95_ms": 1332.429839996621,
    "time_to_first_output_p99_ms": 1332.4311999604106,
    "time_to_first_output_samples": 20,
    "time_to_first_output_stddev_ms": 19.145076485013224,
    "tpot_max_ms": 19.624188160881975,
    "tpot_mean_ci95_high_ms": 14.93321504427994,
    "tpot_mean_ci95_low_ms": 13.610595212083727,
    "tpot_mean_ms": 14.271905128181833,
    "tpot_min_ms": 12.975169820415879,
    "tpot_p25_ms": 13.018094569219855,
    "tpot_p50_ms": 14.652412307138244,
    "tpot_p75_ms": 14.694825751229939,
    "tpot_p90_ms": 14.755010926004115,
    "tpot_p95_ms": 14.824924584626492,
    "tpot_p99_ms": 19.624188160881975,
    "tpot_samples": 20,
    "tpot_stddev_ms": 1.508912169989848,
    "ttft_max_ms": 7103.697716956958,
    "ttft_mean_ci95_high_ms": 4407.346576008413,
    "ttft_mean_ci95_low_ms": 3318.9047358711923,
    "ttft_mean_ms": 3863.1256559398025,
    "ttft_min_ms": 2189.3811889458448,
    "ttft_p25_ms": 2908.5552759934217,
    "ttft_p50_ms": 3657.9408010002226,
    "ttft_p75_ms": 4032.7135069528595,
    "ttft_p90_ms": 5276.299335993826,
    "ttft_p95_ms": 6284.461491974071,
    "ttft_p99_ms": 7103.697716956958,
    "ttft_samples": 20,
    "ttft_stddev_ms": 1241.7499714805015
  },
  "model": "llm.primary",
  "performance_eligible": true,
  "prompt_cache_mode": "unique",
  "request_canaries": false,
  "request_timings": [
    {
      "content_chunks": 42,
      "controlled_output": null,
      "decode_tok_s": 76.44597967925247,
      "e2e_ms": 2812.4596489360556,
      "effective_prefill_tok_s": 4733.350690424071,
      "generation_ms": 1530.4925189120695,
      "mean_inter_token_latency_ms": 13.081132640274099,
      "mean_time_per_output_token_ms": 13.081132640274099,
      "output_token_source": "usage",
      "output_tokens": 118,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6068,
      "reasoning_chunks": 74,
      "request_canary": null,
      "request_index": 0,
      "time_to_first_output_ms": 1281.967130023986,
      "tpot_ms": 13.081132640274099,
      "ttft_ms": 2248.0702749453485,
      "visible_content_truncated": false,
      "visible_generation_ms": 564.3893739907071
    },
    {
      "content_chunks": 43,
      "controlled_output": null,
      "decode_tok_s": 76.96680791379138,
      "e2e_ms": 4016.3873289711773,
      "effective_prefill_tok_s": 4710.62879817782,
      "generation_ms": 2728.448868962005,
      "mean_inter_token_latency_ms": 12.992613661723832,
      "mean_time_per_output_token_ms": 12.992613661723832,
      "output_token_source": "usage",
      "output_tokens": 211,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6067,
      "reasoning_chunks": 157,
      "request_canary": null,
      "request_index": 1,
      "time_to_first_output_ms": 1287.9384600091726,
      "tpot_ms": 12.992613661723832,
      "ttft_ms": 3446.9565480249003,
      "visible_content_truncated": false,
      "visible_generation_ms": 569.4307809462771
    },
    {
      "content_chunks": 44,
      "controlled_output": null,
      "decode_tok_s": 77.07028222679155,
      "e2e_ms": 4977.626550011337,
      "effective_prefill_tok_s": 4694.129932684088,
      "generation_ms": 3684.9482289981097,
      "mean_inter_token_latency_ms": 12.975169820415879,
      "mean_time_per_output_token_ms": 12.975169820415879,
      "output_token_source": "usage",
      "output_tokens": 285,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6068,
      "reasoning_chunks": 230,
      "request_canary": null,
      "request_index": 2,
      "time_to_first_output_ms": 1292.678321013227,
      "tpot_ms": 12.975169820415879,
      "ttft_ms": 4396.986427018419,
      "visible_content_truncated": false,
      "visible_generation_ms": 580.6401229929179
    },
    {
      "content_chunks": 59,
      "controlled_output": null,
      "decode_tok_s": 76.96356158280922,
      "e2e_ms": 4652.42532803677,
      "effective_prefill_tok_s": 4714.898025320386,
      "generation_ms": 3365.228878101334,
      "mean_inter_token_latency_ms": 12.993161691510942,
      "mean_time_per_output_token_ms": 12.993161691510942,
      "output_token_source": "usage",
      "output_tokens": 260,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6069,
      "reasoning_chunks": 191,
      "request_canary": null,
      "request_index": 3,
      "time_to_first_output_ms": 1287.1964499354362,
      "tpot_ms": 12.993161691510942,
      "ttft_ms": 3873.249553027563,
      "visible_content_truncated": false,
      "visible_generation_ms": 779.1757750092074
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 76.81615728652122,
      "e2e_ms": 3546.814661938697,
      "effective_prefill_tok_s": 4736.022505195007,
      "generation_ms": 2265.1484550442547,
      "mean_inter_token_latency_ms": 13.018094569219855,
      "mean_time_per_output_token_ms": 13.018094569219855,
      "output_token_source": "usage",
      "output_tokens": 175,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6070,
      "reasoning_chunks": 121,
      "request_canary": null,
      "request_index": 4,
      "time_to_first_output_ms": 1281.6662068944424,
      "tpot_ms": 13.018094569219855,
      "ttft_ms": 2908.5552759934217,
      "visible_content_truncated": false,
      "visible_generation_ms": 638.2593859452754
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 76.66010654798002,
      "e2e_ms": 3527.9092490673065,
      "effective_prefill_tok_s": 4679.0077393309175,
      "generation_ms": 2230.625650030561,
      "mean_inter_token_latency_ms": 13.044594444623163,
      "mean_time_per_output_token_ms": 13.044594444623163,
      "output_token_source": "usage",
      "output_tokens": 172,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6070,
      "reasoning_chunks": 118,
      "request_canary": null,
      "request_index": 5,
      "time_to_first_output_ms": 1297.2835990367457,
      "tpot_ms": 13.044594444623163,
      "ttft_ms": 2870.9319659974426,
      "visible_content_truncated": false,
      "visible_generation_ms": 656.9772830698639
    },
    {
      "content_chunks": 45,
      "controlled_output": null,
      "decode_tok_s": 76.70721268187934,
      "e2e_ms": 4261.668846942484,
      "effective_prefill_tok_s": 4709.430909144152,
      "generation_ms": 2972.3410879960284,
      "mean_inter_token_latency_ms": 13.036583719280825,
      "mean_time_per_output_token_ms": 13.036583719280825,
      "output_token_source": "usage",
      "output_tokens": 229,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 175,
      "request_canary": null,
      "request_index": 6,
      "time_to_first_output_ms": 1289.327758946456,
      "tpot_ms": 13.036583719280825,
      "ttft_ms": 3657.9408010002226,
      "visible_content_truncated": false,
      "visible_generation_ms": 603.7280459422618
    },
    {
      "content_chunks": 50,
      "controlled_output": null,
      "decode_tok_s": 76.91545674648307,
      "e2e_ms": 3573.050903971307,
      "effective_prefill_tok_s": 4678.595026982243,
      "generation_ms": 2275.22538905032,
      "mean_inter_token_latency_ms": 13.0012879374304,
      "mean_time_per_output_token_ms": 13.0012879374304,
      "output_token_source": "usage",
      "output_tokens": 176,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 116,
      "request_canary": null,
      "request_index": 7,
      "time_to_first_output_ms": 1297.8255149209872,
      "tpot_ms": 13.0012879374304,
      "ttft_ms": 2911.802939954214,
      "visible_content_truncated": false,
      "visible_generation_ms": 661.2479640170932
    },
    {
      "content_chunks": 55,
      "controlled_output": null,
      "decode_tok_s": 50.957522003043046,
      "e2e_ms": 3844.727964955382,
      "effective_prefill_tok_s": 4693.937408008905,
      "generation_ms": 2551.1444609146565,
      "mean_inter_token_latency_ms": 19.624188160881975,
      "mean_time_per_output_token_ms": 19.624188160881975,
      "output_token_source": "usage",
      "output_tokens": 131,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 72,
      "request_canary": null,
      "request_index": 8,
      "time_to_first_output_ms": 1293.5835040407255,
      "tpot_ms": 19.624188160881975,
      "ttft_ms": 2873.0671430239454,
      "visible_content_truncated": false,
      "visible_generation_ms": 971.6608219314367
    },
    {
      "content_chunks": 54,
      "controlled_output": null,
      "decode_tok_s": 67.45396877344011,
      "e2e_ms": 3005.760223022662,
      "effective_prefill_tok_s": 4564.2993873780315,
      "generation_ms": 1675.2164780627936,
      "mean_inter_token_latency_ms": 14.824924584626492,
      "mean_time_per_output_token_ms": 14.824924584626492,
      "output_token_source": "usage",
      "output_tokens": 114,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6073,
      "reasoning_chunks": 58,
      "request_canary": null,
      "request_index": 9,
      "time_to_first_output_ms": 1330.5437449598685,
      "tpot_ms": 14.824924584626492,
      "ttft_ms": 2189.3811889458448,
      "visible_content_truncated": false,
      "visible_generation_ms": 816.3790340768173
    },
    {
      "content_chunks": 42,
      "controlled_output": null,
      "decode_tok_s": 67.98379067005119,
      "e2e_ms": 4670.442081987858,
      "effective_prefill_tok_s": 4612.283161284243,
      "generation_ms": 3353.7406160030514,
      "mean_inter_token_latency_ms": 14.70938866668005,
      "mean_time_per_output_token_ms": 14.70938866668005,
      "output_token_source": "usage",
      "output_tokens": 229,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6073,
      "reasoning_chunks": 178,
      "request_canary": null,
      "request_index": 10,
      "time_to_first_output_ms": 1316.7014659848064,
      "tpot_ms": 14.70938866668005,
      "ttft_ms": 4032.7135069528595,
      "visible_content_truncated": false,
      "visible_generation_ms": 637.7285750349984
    },
    {
      "content_chunks": 59,
      "controlled_output": null,
      "decode_tok_s": 68.21304475004585,
      "e2e_ms": 4745.22425502073,
      "effective_prefill_tok_s": 4619.730342239739,
      "generation_ms": 3430.428899009712,
      "mean_inter_token_latency_ms": 14.659952559870565,
      "mean_time_per_output_token_ms": 14.659952559870565,
      "output_token_source": "usage",
      "output_tokens": 235,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6074,
      "reasoning_chunks": 169,
      "request_canary": null,
      "request_index": 11,
      "time_to_first_output_ms": 1314.7953560110182,
      "tpot_ms": 14.659952559870565,
      "ttft_ms": 3875.227926997468,
      "visible_content_truncated": false,
      "visible_generation_ms": 869.9963280232623
    },
    {
      "content_chunks": 58,
      "controlled_output": null,
      "decode_tok_s": 68.34250823371363,
      "e2e_ms": 7965.677306987345,
      "effective_prefill_tok_s": 4593.749523177495,
      "generation_ms": 6643.010503030382,
      "mean_inter_token_latency_ms": 14.632181724736524,
      "mean_time_per_output_token_ms": 14.632181724736524,
      "output_token_source": "usage",
      "output_tokens": 455,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6076,
      "reasoning_chunks": 385,
      "request_canary": null,
      "request_index": 12,
      "time_to_first_output_ms": 1322.6668039569631,
      "tpot_ms": 14.632181724736524,
      "ttft_ms": 7103.697716956958,
      "visible_content_truncated": false,
      "visible_generation_ms": 861.979590030387
    },
    {
      "content_chunks": 55,
      "controlled_output": null,
      "decode_tok_s": 68.13872286083297,
      "e2e_ms": 7113.6115049012005,
      "effective_prefill_tok_s": 4564.745345894231,
      "generation_ms": 5782.321467995644,
      "mean_inter_token_latency_ms": 14.675942812171684,
      "mean_time_per_output_token_ms": 14.675942812171684,
      "output_token_source": "usage",
      "output_tokens": 395,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6077,
      "reasoning_chunks": 332,
      "request_canary": null,
      "request_index": 13,
      "time_to_first_output_ms": 1331.290036905557,
      "tpot_ms": 14.675942812171684,
      "ttft_ms": 6284.461491974071,
      "visible_content_truncated": false,
      "visible_generation_ms": 829.1500129271299
    },
    {
      "content_chunks": 41,
      "controlled_output": null,
      "decode_tok_s": 68.22103878874331,
      "e2e_ms": 5436.7369239917025,
      "effective_prefill_tok_s": 4560.835861679433,
      "generation_ms": 4104.305724031292,
      "mean_inter_token_latency_ms": 14.658234728683185,
      "mean_time_per_output_token_ms": 14.658234728683185,
      "output_token_source": "usage",
      "output_tokens": 281,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6077,
      "reasoning_chunks": 230,
      "request_canary": null,
      "request_index": 14,
      "time_to_first_output_ms": 1332.4311999604106,
      "tpot_ms": 14.658234728683185,
      "ttft_ms": 4828.489482053556,
      "visible_content_truncated": false,
      "visible_generation_ms": 608.247441938147
    },
    {
      "content_chunks": 43,
      "controlled_output": null,
      "decode_tok_s": 68.24814774785092,
      "e2e_ms": 4142.896914971061,
      "effective_prefill_tok_s": 4571.18359913417,
      "generation_ms": 2813.263162970543,
      "mean_inter_token_latency_ms": 14.652412307138244,
      "mean_time_per_output_token_ms": 14.652412307138244,
      "output_token_source": "usage",
      "output_tokens": 193,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 145,
      "request_canary": null,
      "request_index": 15,
      "time_to_first_output_ms": 1329.6337520005181,
      "tpot_ms": 14.652412307138244,
      "ttft_ms": 3508.441500016488,
      "visible_content_truncated": false,
      "visible_generation_ms": 634.4554149545729
    },
    {
      "content_chunks": 59,
      "controlled_output": null,
      "decode_tok_s": 68.1396281823721,
      "e2e_ms": 6152.306975913234,
      "effective_prefill_tok_s": 4591.438479390795,
      "generation_ms": 4828.3210339723155,
      "mean_inter_token_latency_ms": 14.675747823624059,
      "mean_time_per_output_token_ms": 14.675747823624059,
      "output_token_source": "usage",
      "output_tokens": 330,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6079,
      "reasoning_chunks": 264,
      "request_canary": null,
      "request_index": 16,
      "time_to_first_output_ms": 1323.9859419409186,
      "tpot_ms": 14.675747823624059,
      "ttft_ms": 5276.299335993826,
      "visible_content_truncated": false,
      "visible_generation_ms": 876.0076399194077
    },
    {
      "content_chunks": 53,
      "controlled_output": null,
      "decode_tok_s": 67.77358586957112,
      "e2e_ms": 4312.904532067478,
      "effective_prefill_tok_s": 4614.319839754458,
      "generation_ms": 2995.267217978835,
      "mean_inter_token_latency_ms": 14.755010926004115,
      "mean_time_per_output_token_ms": 14.755010926004115,
      "output_token_source": "usage",
      "output_tokens": 204,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6080,
      "reasoning_chunks": 143,
      "request_canary": null,
      "request_index": 17,
      "time_to_first_output_ms": 1317.6373140886426,
      "tpot_ms": 14.755010926004115,
      "ttft_ms": 3512.413377990015,
      "visible_content_truncated": false,
      "visible_generation_ms": 800.4911540774629
    },
    {
      "content_chunks": 42,
      "controlled_output": null,
      "decode_tok_s": 68.05116419405662,
      "e2e_ms": 4344.8691189987585,
      "effective_prefill_tok_s": 4562.341533881751,
      "generation_ms": 3012.4392790021375,
      "mean_inter_token_latency_ms": 14.694825751229939,
      "mean_time_per_output_token_ms": 14.694825751229939,
      "output_token_source": "usage",
      "output_tokens": 206,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6079,
      "reasoning_chunks": 157,
      "request_canary": null,
      "request_index": 18,
      "time_to_first_output_ms": 1332.429839996621,
      "tpot_ms": 14.694825751229939,
      "ttft_ms": 3716.104042949155,
      "visible_content_truncated": false,
      "visible_generation_ms": 628.7650760496035
    },
    {
      "content_chunks": 43,
      "controlled_output": null,
      "decode_tok_s": 67.87643270013697,
      "e2e_ms": 4392.231123987585,
      "effective_prefill_tok_s": 4578.11497537042,
      "generation_ms": 3064.3920389702544,
      "mean_inter_token_latency_ms": 14.732654033510839,
      "mean_time_per_output_token_ms": 14.732654033510839,
      "output_token_source": "usage",
      "output_tokens": 209,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6079,
      "reasoning_chunks": 158,
      "request_canary": null,
      "request_index": 19,
      "time_to_first_output_ms": 1327.839085017331,
      "tpot_ms": 14.732654033510839,
      "ttft_ms": 3747.722618980333,
      "visible_content_truncated": false,
      "visible_generation_ms": 644.5085050072521
    }
  ],
  "requests": 20,
  "response_words": 0,
  "responses_received": 20,
  "run_id": "benchmark-20261017T005017Z",
  "sampling": {
    "temperature": {
      "effective_request": 0.6,
      "requested": 0.6,
      "sent": true
    },
    "top_p": {
      "effective_request": 0.95,
      "requested": 0.95,
      "sent": true
    }
  },
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "chat_template_kwargs": null,
    "force_nonempty_content": false,
    "no_thinking": false,
    "reasoning_effort": "max",
    "shared_prefix_burst": false,
    "thinking_mode": "default"
  },
  "shared_prefix_tokens": 0,
  "source_recipe": null,
  "started_at": "2026-10-17T00:50:17Z",
  "started_at_unix_ns": 1792198217703607339,
  "tier": null,
  "timestamp_precision": "nanosecond-unix-fields; ISO labels are second precision",
  "timing_methodology": {
    "artifact_timestamps": "time.time_ns Unix timestamps; independent artifacts may align them only when the operator supplied the same explicit clock_domain_id",
    "clock": "client time.perf_counter",
    "controlled_output": "--response-words adherence is recorded; observe preserves historical behavior while strict rejects non-exact or truncated completions; visible response capture is capped at 8192 characters",
    "decode": "usage completion tokens after the first token divided by client-observed generation time; completion tokens may include reasoning tokens",
    "distributions": "request-level mean, sample standard deviation, normal-approximation 95% confidence interval for the mean, min, p25, p50, p75, p90, p95, p99, and max using the nearest-rank percentile method",
    "effective_prefill": "usage.prompt_tokens divided by client-observed time to first output; includes queueing, scheduling, prefill, and first-token work",
    "generation": "client-observed E2E minus time to first output",
    "mean_inter_token_latency": "legacy key for a TPOT proxy: client-observed generation time divided by completion tokens after the first token; not raw per-token ITL",
    "p99": "numeric nearest-rank p99 is always retained for compatibility; it is descriptive only below 100 eligible samples",
    "request_canaries": "when enabled, each response must begin with its own synthetic marker and contain no other ANVIL_REQ marker; bounded/truncated captures fail closed and are excluded from performance metrics",
    "time_to_first_output": "request start through the first non-empty streamed reasoning or content delta",
    "tpot": "client-observed mean time per output token after the first output; identical to the legacy mean_inter_token_latency key",
    "ttft": "request start through the first non-empty streamed content delta"
  },
  "unique_prefix_namespace": "capacity-v1:seed-request-index",
  "wall_clock_ms": 91509.01760498527
}