{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 32,
  "concurrency": 16,
  "context_distribution": {
    "4096": 32
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-k5-bt4096",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-30T00:23:11Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 1048576,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 429,
    "content_chunks_s": null,
    "decode_tok_s_p50": 4.350258133442958,
    "decode_tok_s_p95": 39.85236293627566,
    "e2e_p50_ms": 16322.673499998928,
    "e2e_p95_ms": 31949.977299998864,
    "effective_prefill_tok_s_p50": 452.2277733555559,
    "effective_prefill_tok_s_p95": 929.0402907362825,
    "generation_p50_ms": 9632.080899995344,
    "generation_p95_ms": 19098.592599999392,
    "mean_inter_token_latency_ms_p50": 219.20179500011727,
    "mean_inter_token_latency_ms_p95": 561.9936275861933,
    "output_token_sources": {
      "usage": 32
    },
    "output_tokens": 1239,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 32,
    "prompt_tokens": 94976,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 12,
    "throughput_tok_s": 31.9945219180999,
    "time_to_first_output_p50_ms": 6528.256200006581,
    "time_to_first_output_p95_ms": 22233.221099995717,
    "ttft_p50_ms": 6563.065199996345,
    "ttft_p95_ms": 22317.897200002335,
    "ttft_samples": 32
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision-bt4096",
  "request_timings": [
    {
      "content_chunks": 9,
      "decode_tok_s": 1.6159150148986994,
      "e2e_ms": 19334.751499998674,
      "effective_prefill_tok_s": 581.809091852746,
      "generation_ms": 14233.421799995995,
      "mean_inter_token_latency_ms": 618.8444260867824,
      "output_token_source": "usage",
      "output_tokens": 24,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 5101.32970000268,
      "ttft_ms": 5101.330299999972,
      "visible_generation_ms": 14233.421199998702
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 3.0107720544595944,
      "e2e_ms": 31949.977299998864,
      "effective_prefill_tok_s": 132.9874441033579,
      "generation_ms": 9632.080899995344,
      "mean_inter_token_latency_ms": 332.14072068949463,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 22317.89640000352,
      "ttft_ms": 22317.897200002335,
      "visible_generation_ms": 9632.080099996529
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 2.6988622413032566,
      "e2e_ms": 22318.205800002033,
      "effective_prefill_tok_s": 416.4667705680637,
      "generation_ms": 15191.58679999964,
      "mean_inter_token_latency_ms": 370.5265073170644,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 7126.619000002393,
      "ttft_ms": 7126.619400005438,
      "visible_generation_ms": 15191.586399996595
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 4.248389357880875,
      "e2e_ms": 26115.63659999956,
      "effective_prefill_tok_s": 180.26203151274976,
      "generation_ms": 9650.716199997987,
      "mean_inter_token_latency_ms": 235.38332195117042,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 16464.920400001574,
      "ttft_ms": 16464.92109999963,
      "visible_generation_ms": 9650.715499999933
    },
    {
      "content_chunks": 8,
      "decode_tok_s": 4.661494151150104,
      "e2e_ms": 24267.404400001396,
      "effective_prefill_tok_s": 153.51699707554522,
      "generation_ms": 4934.04030000238,
      "mean_inter_token_latency_ms": 214.52349130445128,
      "output_token_source": "usage",
      "output_tokens": 24,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 19333.364099999017,
      "ttft_ms": 19333.364599995548,
      "visible_generation_ms": 4934.0398000058485
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 2.5827832367741643,
      "e2e_ms": 24102.372500005004,
      "effective_prefill_tok_s": 929.0402907362825,
      "generation_ms": 20907.677900002454,
      "mean_inter_token_latency_ms": 387.1792203704158,
      "output_token_source": "usage",
      "output_tokens": 55,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 3194.694600002549,
      "ttft_ms": 3194.6954000013648,
      "visible_generation_ms": 20907.67710000364
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 4.562006437946665,
      "e2e_ms": 27922.797800005355,
      "effective_prefill_tok_s": 154.9487056092526,
      "generation_ms": 8768.07180000469,
      "mean_inter_token_latency_ms": 219.20179500011727,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 19154.726000000664,
      "ttft_ms": 19154.72640000371,
      "visible_generation_ms": 8768.071400001645
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 2.47177899990494,
      "e2e_ms": 20390.30659999844,
      "effective_prefill_tok_s": 591.6164402316143,
      "generation_ms": 15373.542699999234,
      "mean_inter_token_latency_ms": 404.5669131578746,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 5016.763899999205,
      "ttft_ms": 7042.7009999984875,
      "visible_generation_ms": 13347.605599999952
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 1.7793796066604501,
      "e2e_ms": 20388.95649999904,
      "effective_prefill_tok_s": 725.4699318257254,
      "generation_ms": 16297.815199999604,
      "mean_inter_token_latency_ms": 561.9936275861933,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 4091.1412999994354,
      "ttft_ms": 4091.142299999774,
      "visible_generation_ms": 16297.814199999266
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 3.9451848950867583,
      "e2e_ms": 31865.215799996804,
      "effective_prefill_tok_s": 133.49392724748157,
      "generation_ms": 9631.994700001087,
      "mean_inter_token_latency_ms": 253.4735447368707,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 9,
      "time_to_first_output_ms": 22233.221099995717,
      "ttft_ms": 24077.716400002828,
      "visible_generation_ms": 7787.499399993976
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 4.8629524419681465,
      "e2e_ms": 24183.42300000222,
      "effective_prefill_tok_s": 162.89820405606542,
      "generation_ms": 5963.4554000003845,
      "mean_inter_token_latency_ms": 205.63639310346153,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 18219.967600001837,
      "ttft_ms": 18219.96809999837,
      "visible_generation_ms": 5963.454900003853
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.141145863136177,
      "e2e_ms": 25113.278199998604,
      "effective_prefill_tok_s": 192.0523968634138,
      "generation_ms": 9659.16229999857,
      "mean_inter_token_latency_ms": 241.47905749996426,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 11,
      "time_to_first_output_ms": 15454.115900000033,
      "ttft_ms": 15454.116399996565,
      "visible_generation_ms": 9659.161800002039
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 2.6179939562667873,
      "e2e_ms": 20274.197099999583,
      "effective_prefill_tok_s": 2524.65859053748,
      "generation_ms": 19098.592599999392,
      "mean_inter_token_latency_ms": 381.97185199998785,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 1175.6045000001905,
      "ttft_ms": 1175.6056000012904,
      "visible_generation_ms": 19098.591499998292
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.729175355344455,
      "e2e_ms": 25112.16389999754,
      "effective_prefill_tok_s": 204.13362193482996,
      "generation_ms": 10572.66779999918,
      "mean_inter_token_latency_ms": 211.45335599998361,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 14539.49609999836,
      "ttft_ms": 14539.49649999413,
      "visible_generation_ms": 10572.66740000341
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.350258133442958,
      "e2e_ms": 26950.924099997792,
      "effective_prefill_tok_s": 171.59714960483635,
      "generation_ms": 9654.599499997857,
      "mean_inter_token_latency_ms": 229.87141666661566,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 14,
      "time_to_first_output_ms": 17296.324599999934,
      "ttft_ms": 17296.324999995704,
      "visible_generation_ms": 9654.599100002088
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 3.573866432296756,
      "e2e_ms": 33817.24060000124,
      "effective_prefill_tok_s": 145.58719354611515,
      "generation_ms": 13430.832099998952,
      "mean_inter_token_latency_ms": 279.8090020833115,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 20386.40850000229,
      "ttft_ms": 20386.40889999806,
      "visible_generation_ms": 13430.831700003182
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 3.4500906937513434,
      "e2e_ms": 14683.012000001327,
      "effective_prefill_tok_s": 614.722020755296,
      "generation_ms": 9854.813400001149,
      "mean_inter_token_latency_ms": 289.84745294121024,
      "output_token_source": "usage",
      "output_tokens": 35,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 16,
      "time_to_first_output_ms": 4828.198600000178,
      "ttft_ms": 4828.199000003224,
      "visible_generation_ms": 9854.812999998103
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.10066675559993,
      "e2e_ms": 15507.106299999577,
      "effective_prefill_tok_s": 780.7048481140329,
      "generation_ms": 11705.41349999985,
      "mean_inter_token_latency_ms": 243.86278124999686,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 17,
      "time_to_first_output_ms": 3801.692799999728,
      "ttft_ms": 4838.238600001205,
      "visible_generation_ms": 10668.867699998373
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.71696028496782,
      "e2e_ms": 15394.436900001892,
      "effective_prefill_tok_s": 526.0178756044327,
      "generation_ms": 9752.043100001174,
      "mean_inter_token_latency_ms": 212.00093695654726,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 18,
      "time_to_first_output_ms": 5642.393800000718,
      "ttft_ms": 5642.394399998011,
      "visible_generation_ms": 9752.042500003881
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.098524806694796,
      "e2e_ms": 16322.673499998928,
      "effective_prefill_tok_s": 452.2277733555559,
      "generation_ms": 9759.609100001398,
      "mean_inter_token_latency_ms": 243.99022750003496,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 19,
      "time_to_first_output_ms": 6563.064399997529,
      "ttft_ms": 6563.065199996345,
      "visible_generation_ms": 9759.608300002583
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 4.262396076215625,
      "e2e_ms": 15763.234500002,
      "effective_prefill_tok_s": 522.9944888222542,
      "generation_ms": 10088.222500002303,
      "mean_inter_token_latency_ms": 234.6098255814489,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 20,
      "time_to_first_output_ms": 5675.011999999697,
      "ttft_ms": 5675.012499996228,
      "visible_generation_ms": 10088.222000005771
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 5.322414845211148,
      "e2e_ms": 10782.819899999595,
      "effective_prefill_tok_s": 774.7146098071028,
      "generation_ms": 6951.731699999073,
      "mean_inter_token_latency_ms": 187.88464054051548,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 21,
      "time_to_first_output_ms": 3831.088200000522,
      "ttft_ms": 5700.0499999994645,
      "visible_generation_ms": 5082.76990000013
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 5.095980936553876,
      "e2e_ms": 13639.847200000077,
      "effective_prefill_tok_s": 530.5408484996973,
      "generation_ms": 8045.556000004581,
      "mean_inter_token_latency_ms": 196.23307317084343,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 22,
      "time_to_first_output_ms": 5594.291199995496,
      "ttft_ms": 5594.291599998542,
      "visible_generation_ms": 8045.555600001535
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 4.832829833010815,
      "e2e_ms": 12528.35590000177,
      "effective_prefill_tok_s": 454.6756125420806,
      "generation_ms": 6000.6251000013435,
      "mean_inter_token_latency_ms": 206.91810689659806,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 23,
      "time_to_first_output_ms": 6527.730800000427,
      "ttft_ms": 6527.731500005757,
      "visible_generation_ms": 6000.624399996013
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 5.109167499554112,
      "e2e_ms": 12709.43090000219,
      "effective_prefill_tok_s": 446.85966808754574,
      "generation_ms": 6067.52470000356,
      "mean_inter_token_latency_ms": 195.72660322592128,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 24,
      "time_to_first_output_ms": 6641.90619999863,
      "ttft_ms": 6641.9066000016755,
      "visible_generation_ms": 6067.524300000514
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 5.507490002823907,
      "e2e_ms": 12520.09640000324,
      "effective_prefill_tok_s": 454.6390198345782,
      "generation_ms": 5991.840199996659,
      "mean_inter_token_latency_ms": 181.5709151514139,
      "output_token_source": "usage",
      "output_tokens": 34,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 25,
      "time_to_first_output_ms": 6528.256200006581,
      "ttft_ms": 6528.256600002351,
      "visible_generation_ms": 5991.839800000889
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 9.173818851803818,
      "e2e_ms": 12257.661700001336,
      "effective_prefill_tok_s": 381.0785022355959,
      "generation_ms": 4469.240200000058,
      "mean_inter_token_latency_ms": 109.00585853658679,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 26,
      "time_to_first_output_ms": 7788.421500001277,
      "ttft_ms": 7788.4221999993315,
      "visible_generation_ms": 4469.239500002004
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 8.275743955634258,
      "e2e_ms": 11571.201099999598,
      "effective_prefill_tok_s": 425.24725627734426,
      "generation_ms": 4591.732199995022,
      "mean_inter_token_latency_ms": 120.83505789460584,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 27,
      "time_to_first_output_ms": 6979.468900004576,
      "ttft_ms": 6979.469099998823,
      "visible_generation_ms": 4591.732000000775
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 12.001707494538698,
      "e2e_ms": 10456.261299994367,
      "effective_prefill_tok_s": 376.9704820557305,
      "generation_ms": 2582.9658000002382,
      "mean_inter_token_latency_ms": 83.32147741936252,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 28,
      "time_to_first_output_ms": 7873.2954999941285,
      "ttft_ms": 7873.295999997936,
      "visible_generation_ms": 2582.965299996431
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 39.85236293627566,
      "e2e_ms": 6773.18959999684,
      "effective_prefill_tok_s": 514.4306640889965,
      "generation_ms": 1003.7045999997645,
      "mean_inter_token_latency_ms": 25.09261499999411,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 29,
      "time_to_first_output_ms": 5769.484999997076,
      "ttft_ms": 5769.485600001644,
      "visible_generation_ms": 1003.7039999951958
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 17.524997713251505,
      "e2e_ms": 6615.712300001178,
      "effective_prefill_tok_s": 612.3614694654248,
      "generation_ms": 1768.9018000019132,
      "mean_inter_token_latency_ms": 57.06134838715849,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 30,
      "time_to_first_output_ms": 4846.810499999265,
      "ttft_ms": 4846.8109000023105,
      "visible_generation_ms": 1768.9013999988674
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 41.043332843339975,
      "e2e_ms": 4611.4084999935585,
      "effective_prefill_tok_s": 760.0827097016978,
      "generation_ms": 706.5702999971109,
      "mean_inter_token_latency_ms": 24.36449310334865,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 31,
      "time_to_first_output_ms": 3904.8381999964477,
      "ttft_ms": 3904.838499998732,
      "visible_generation_ms": 706.5699999948265
    }
  ],
  "requests": 32,
  "run_id": "benchmark-20260830T002232Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": "low",
    "shared_prefix_burst": false,
    "thinking_mode": "default"
  },
  "source_recipe": null,
  "started_at": "2026-08-30T00:22:32Z",
  "tier": null,
  "timing_methodology": {
    "clock": "client time.perf_counter",
    "decode": "usage completion tokens after the first token divided by client-observed generation time; completion tokens may include reasoning tokens",
    "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": "client-observed generation time divided by completion tokens after the first token; not a raw per-token timestamp trace",
    "time_to_first_output": "request start through the first non-empty streamed reasoning or content delta",
    "ttft": "request start through the first non-empty streamed content delta"
  },
  "wall_clock_ms": 38725.37939999893
}
