{
  "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",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-30T00:08:08Z",
  "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": 430,
    "content_chunks_s": null,
    "decode_tok_s_p50": 4.076155828683055,
    "decode_tok_s_p95": 5.257554764455183,
    "e2e_p50_ms": 20172.93079999945,
    "e2e_p95_ms": 37485.56069999904,
    "effective_prefill_tok_s_p50": 243.2827026891941,
    "effective_prefill_tok_s_p95": 533.707312068629,
    "generation_p50_ms": 9433.184900000924,
    "generation_p95_ms": 16820.206999997026,
    "mean_inter_token_latency_ms_p50": 243.5680028572928,
    "mean_inter_token_latency_ms_p95": 370.6187161292145,
    "output_token_sources": {
      "usage": 32
    },
    "output_tokens": 1334,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 32,
    "prompt_tokens": 94976,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 47,
    "throughput_tok_s": 26.348208498202798,
    "time_to_first_output_p50_ms": 11815.351099998225,
    "time_to_first_output_p95_ms": 28055.23979999998,
    "ttft_p50_ms": 12199.799000001804,
    "ttft_p95_ms": 29305.493699997896,
    "ttft_samples": 32
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 10,
      "decode_tok_s": 4.853789828313575,
      "e2e_ms": 15727.907399996184,
      "effective_prefill_tok_s": 357.11610265629224,
      "generation_ms": 7416.884799997206,
      "mean_inter_token_latency_ms": 206.02457777770016,
      "output_token_source": "usage",
      "output_tokens": 37,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 0,
      "time_to_first_output_ms": 8311.022599998978,
      "ttft_ms": 10067.314899999474,
      "visible_generation_ms": 5660.59249999671
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.346358142518333,
      "e2e_ms": 22022.69299999898,
      "effective_prefill_tok_s": 235.7518638873951,
      "generation_ms": 9433.184900000924,
      "mean_inter_token_latency_ms": 230.07768048782742,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 12589.508099998056,
      "ttft_ms": 12589.508800003387,
      "visible_generation_ms": 9433.184199995594
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 3.3638263120643765,
      "e2e_ms": 31318.08639999508,
      "effective_prefill_tok_s": 235.75362040478245,
      "generation_ms": 18728.67209999822,
      "mean_inter_token_latency_ms": 297.2805095237813,
      "output_token_source": "usage",
      "output_tokens": 64,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 12589.41429999686,
      "ttft_ms": 13860.473900000216,
      "visible_generation_ms": 17457.612499994866
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 2.975174149809492,
      "e2e_ms": 28162.919099995634,
      "effective_prefill_tok_s": 188.72396800689523,
      "generation_ms": 12436.246799996297,
      "mean_inter_token_latency_ms": 336.1147783782783,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 3,
      "time_to_first_output_ms": 15726.672299999336,
      "ttft_ms": 16966.603699998814,
      "visible_generation_ms": 11196.31539999682
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 2.8659695841082455,
      "e2e_ms": 29412.829699998838,
      "effective_prefill_tok_s": 196.4648844768533,
      "generation_ms": 14305.804299998272,
      "mean_inter_token_latency_ms": 348.9220560975188,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 15107.025400000566,
      "ttft_ms": 15107.026300000143,
      "visible_generation_ms": 14305.803399998695
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 3.124705691783503,
      "e2e_ms": 30003.830699999526,
      "effective_prefill_tok_s": 157.8490341835377,
      "generation_ms": 11201.054899996961,
      "mean_inter_token_latency_ms": 320.03013999991316,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 18802.775800002564,
      "ttft_ms": 18802.77650000062,
      "visible_generation_ms": 11201.054199998907
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.50344713718865,
      "e2e_ms": 20038.658299999952,
      "effective_prefill_tok_s": 279.4894334488952,
      "generation_ms": 9419.294399995124,
      "mean_inter_token_latency_ms": 285.43316363621585,
      "output_token_source": "usage",
      "output_tokens": 34,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 10619.363900004828,
      "ttft_ms": 10619.36710000009,
      "visible_generation_ms": 9419.291199999861
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 2.6795607128168784,
      "e2e_ms": 29325.305799997295,
      "effective_prefill_tok_s": 182.49495481078569,
      "generation_ms": 13061.84249999933,
      "mean_inter_token_latency_ms": 373.1954999999808,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 16263.463299997966,
      "ttft_ms": 17505.81070000044,
      "visible_generation_ms": 11819.495099996857
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 2.969391305758956,
      "e2e_ms": 31841.633700001694,
      "effective_prefill_tok_s": 153.13834937965424,
      "generation_ms": 12460.466200005612,
      "mean_inter_token_latency_ms": 336.76935675690845,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 8,
      "time_to_first_output_ms": 19381.16749999608,
      "ttft_ms": 20674.344400002155,
      "visible_generation_ms": 11167.289299999538
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 4.655281090529946,
      "e2e_ms": 18117.7119999993,
      "effective_prefill_tok_s": 280.0166689167159,
      "generation_ms": 7518.342999996094,
      "mean_inter_token_latency_ms": 214.80979999988838,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 9,
      "time_to_first_output_ms": 10599.369000003207,
      "ttft_ms": 11850.48590000224,
      "visible_generation_ms": 6267.22609999706
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 2.9749409626992294,
      "e2e_ms": 29925.58210000425,
      "effective_prefill_tok_s": 169.7128784598672,
      "generation_ms": 12437.221599997429,
      "mean_inter_token_latency_ms": 336.1411243242548,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 10,
      "time_to_first_output_ms": 17488.36050000682,
      "ttft_ms": 18724.89250000217,
      "visible_generation_ms": 11200.689600002079
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 2.9062098915990298,
      "e2e_ms": 29926.037199998973,
      "effective_prefill_tok_s": 215.79536259541345,
      "generation_ms": 16172.266199995647,
      "mean_inter_token_latency_ms": 344.09077021267336,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 11,
      "time_to_first_output_ms": 13753.771000003326,
      "ttft_ms": 15001.46610000229,
      "visible_generation_ms": 14924.571099996683
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 3.2698765241123207,
      "e2e_ms": 37485.051299998304,
      "effective_prefill_tok_s": 143.6255679893904,
      "generation_ms": 16820.206999997026,
      "mean_inter_token_latency_ms": 305.8219454544914,
      "output_token_source": "usage",
      "output_tokens": 56,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 12,
      "time_to_first_output_ms": 20664.844300001278,
      "ttft_ms": 21915.168699997594,
      "visible_generation_ms": 15569.88260000071
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 4.03433561919355,
      "e2e_ms": 37485.56069999904,
      "effective_prefill_tok_s": 101.27696605001503,
      "generation_ms": 8179.785499996797,
      "mean_inter_token_latency_ms": 247.8722878786908,
      "output_token_source": "usage",
      "output_tokens": 34,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 29305.775200002245,
      "ttft_ms": 29305.7759000003,
      "visible_generation_ms": 8179.7847999987425
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 4.984339999097685,
      "e2e_ms": 37484.773099997255,
      "effective_prefill_tok_s": 105.79128965420577,
      "generation_ms": 9429.533299997274,
      "mean_inter_token_latency_ms": 200.62836808504838,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 14,
      "time_to_first_output_ms": 28055.23979999998,
      "ttft_ms": 29305.493699997896,
      "visible_generation_ms": 8179.279399999359
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 3.1553070403981036,
      "e2e_ms": 38098.025700004655,
      "effective_prefill_tok_s": 139.33665106665174,
      "generation_ms": 16797.097499998927,
      "mean_inter_token_latency_ms": 316.92636792450804,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 15,
      "time_to_first_output_ms": 21300.92820000573,
      "ttft_ms": 28055.106500003603,
      "visible_generation_ms": 10042.919200001052
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 4.6295053188381035,
      "e2e_ms": 21865.278999997827,
      "effective_prefill_tok_s": 207.47879983305506,
      "generation_ms": 7560.203000000911,
      "mean_inter_token_latency_ms": 216.00580000002603,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 16,
      "time_to_first_output_ms": 14305.075999996916,
      "ttft_ms": 15590.358000001288,
      "visible_generation_ms": 6274.920999996539
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.984737345580779,
      "e2e_ms": 23063.716000004206,
      "effective_prefill_tok_s": 238.752948303318,
      "generation_ms": 10632.45590000588,
      "mean_inter_token_latency_ms": 200.61237547180906,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 17,
      "time_to_first_output_ms": 12431.260099998326,
      "ttft_ms": 13715.698899999552,
      "visible_generation_ms": 9348.017100004654
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 4.559800327080011,
      "e2e_ms": 19929.74150000373,
      "effective_prefill_tok_s": 251.1986292138577,
      "generation_ms": 8114.3904000055045,
      "mean_inter_token_latency_ms": 219.30784864879743,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 2,
      "request_index": 18,
      "time_to_first_output_ms": 11815.351099998225,
      "ttft_ms": 13092.087099998025,
      "visible_generation_ms": 6837.654400005704
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 4.736977433226185,
      "e2e_ms": 21748.98600000597,
      "effective_prefill_tok_s": 265.1483884016584,
      "generation_ms": 10555.254000006244,
      "mean_inter_token_latency_ms": 211.10508000012487,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 19,
      "time_to_first_output_ms": 11193.731999999727,
      "ttft_ms": 11193.732500003534,
      "visible_generation_ms": 10555.253500002436
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 4.9332769996001264,
      "e2e_ms": 14990.468300005887,
      "effective_prefill_tok_s": 523.8228656263918,
      "generation_ms": 9324.43079999939,
      "mean_inter_token_latency_ms": 202.7050173912911,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 20,
      "time_to_first_output_ms": 5666.037500006496,
      "ttft_ms": 6921.005500000319,
      "visible_generation_ms": 8069.462800005567
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 4.707890975354156,
      "e2e_ms": 14357.561200005875,
      "effective_prefill_tok_s": 472.1598756243034,
      "generation_ms": 8071.55480000074,
      "mean_inter_token_latency_ms": 212.40933684212473,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 21,
      "time_to_first_output_ms": 6286.006400005135,
      "ttft_ms": 7535.284199999296,
      "visible_generation_ms": 6822.277000006579
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.374171209663756,
      "e2e_ms": 14358.015300000261,
      "effective_prefill_tok_s": 523.3962995426145,
      "generation_ms": 8687.360000003537,
      "mean_inter_token_latency_ms": 228.61473684219834,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 22,
      "time_to_first_output_ms": 5670.655299996724,
      "ttft_ms": 5670.6559999947785,
      "visible_generation_ms": 8687.359300005483
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 3.274349359885508,
      "e2e_ms": 20092.376200002036,
      "effective_prefill_tok_s": 315.6362329423481,
      "generation_ms": 10689.146500000788,
      "mean_inter_token_latency_ms": 305.40418571430826,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 23,
      "time_to_first_output_ms": 9403.229700001248,
      "ttft_ms": 9403.230500000063,
      "visible_generation_ms": 10689.145700001973
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 2.698190772565719,
      "e2e_ms": 19662.621900002705,
      "effective_prefill_tok_s": 363.1273224841218,
      "generation_ms": 11489.180200005649,
      "mean_inter_token_latency_ms": 370.6187161292145,
      "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": 8173.441699997056,
      "ttft_ms": 8173.442100000102,
      "visible_generation_ms": 11489.179800002603
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 3.696435778940387,
      "e2e_ms": 20172.93079999945,
      "effective_prefill_tok_s": 429.09284163826635,
      "generation_ms": 13256.012799996824,
      "mean_inter_token_latency_ms": 270.53087346932296,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 25,
      "time_to_first_output_ms": 6916.9180000026245,
      "ttft_ms": 6916.918300004909,
      "visible_generation_ms": 13256.01249999454
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.076155828683055,
      "e2e_ms": 19039.220799997565,
      "effective_prefill_tok_s": 339.76712375065324,
      "generation_ms": 10303.82589999499,
      "mean_inter_token_latency_ms": 245.3291880951188,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 26,
      "time_to_first_output_ms": 8735.394900002575,
      "ttft_ms": 8735.395499999868,
      "visible_generation_ms": 10303.825299997698
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 5.16058759482584,
      "e2e_ms": 18649.10389999568,
      "effective_prefill_tok_s": 317.5065890908707,
      "generation_ms": 9301.266399997985,
      "mean_inter_token_latency_ms": 193.77638333329136,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 27,
      "time_to_first_output_ms": 9347.837499997695,
      "ttft_ms": 9347.837699999218,
      "visible_generation_ms": 9301.266199996462
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 4.105629591198409,
      "e2e_ms": 12836.787899999763,
      "effective_prefill_tok_s": 688.3264062380405,
      "generation_ms": 8524.880100005248,
      "mean_inter_token_latency_ms": 243.5680028572928,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 28,
      "time_to_first_output_ms": 4311.907799994515,
      "ttft_ms": 4311.908499999845,
      "visible_generation_ms": 8524.879399999918
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 40.71372111084176,
      "e2e_ms": 13034.897999998066,
      "effective_prefill_tok_s": 243.2827026891941,
      "generation_ms": 835.0992999985465,
      "mean_inter_token_latency_ms": 24.56174411760431,
      "output_token_source": "usage",
      "output_tokens": 35,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 29,
      "time_to_first_output_ms": 12199.79869999952,
      "ttft_ms": 12199.799000001804,
      "visible_generation_ms": 835.0989999962621
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 4.261184258157853,
      "e2e_ms": 12836.073400001624,
      "effective_prefill_tok_s": 533.707312068629,
      "generation_ms": 7274.972899998829,
      "mean_inter_token_latency_ms": 234.67654516125256,
      "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": 5561.100500002794,
      "ttft_ms": 5561.100699997041,
      "visible_generation_ms": 7274.972700004582
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 5.257554764455183,
      "e2e_ms": 12835.986699996283,
      "effective_prefill_tok_s": 480.3444289662766,
      "generation_ms": 6657.0871000003535,
      "mean_inter_token_latency_ms": 190.20248857143866,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 31,
      "time_to_first_output_ms": 6178.89959999593,
      "ttft_ms": 6178.899799997453,
      "visible_generation_ms": 6657.086899998831
    }
  ],
  "requests": 32,
  "run_id": "benchmark-20260830T000717Z",
  "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": "2026-08-30T00:07:17Z",
  "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": 50629.62819999666
}
