{
  "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-k3",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T23:59:04Z",
  "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": 480,
    "content_chunks_s": null,
    "decode_tok_s_p50": 4.439465928697213,
    "decode_tok_s_p95": 49.11521857279371,
    "e2e_p50_ms": 14203.404200001387,
    "e2e_p95_ms": 24845.96590000001,
    "effective_prefill_tok_s_p50": 473.79676414361546,
    "effective_prefill_tok_s_p95": 981.8628299901037,
    "generation_p50_ms": 8188.883300004818,
    "generation_p95_ms": 12541.788700000325,
    "mean_inter_token_latency_ms_p50": 222.1605741935726,
    "mean_inter_token_latency_ms_p95": 285.31233636352806,
    "output_token_sources": {
      "usage": 32
    },
    "output_tokens": 1303,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 32,
    "prompt_tokens": 94976,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 21,
    "throughput_tok_s": 41.64443605916546,
    "time_to_first_output_p50_ms": 5645.024600002216,
    "time_to_first_output_p95_ms": 14831.717200002458,
    "ttft_p50_ms": 6264.289900005679,
    "ttft_p95_ms": 14831.717700006266,
    "ttft_samples": 32
  },
  "model": "glm53-flash-exl3-k3-dflash2-k3-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 4.343879892419879,
      "e2e_ms": 15554.71719999332,
      "effective_prefill_tok_s": 343.1829261314331,
      "generation_ms": 6906.268299993826,
      "mean_inter_token_latency_ms": 230.20894333312754,
      "output_token_source": "usage",
      "output_tokens": 31,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 8648.448899999494,
      "ttft_ms": 8648.44909999374,
      "visible_generation_ms": 6906.268099999579
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 4.090855669416142,
      "e2e_ms": 13045.350899999903,
      "effective_prefill_tok_s": 981.8068673400912,
      "generation_ms": 10022.35309999378,
      "mean_inter_token_latency_ms": 244.44763658521413,
      "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": 3022.9978000061237,
      "ttft_ms": 3022.9988000064623,
      "visible_generation_ms": 10022.35209999344
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 4.106229245576282,
      "e2e_ms": 11789.99310000654,
      "effective_prefill_tok_s": 981.8628299901037,
      "generation_ms": 8767.167600002722,
      "mean_inter_token_latency_ms": 243.53243333340893,
      "output_token_source": "usage",
      "output_tokens": 37,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 3022.825500003819,
      "ttft_ms": 4899.030300002778,
      "visible_generation_ms": 6890.962800003763
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 4.364076452973208,
      "e2e_ms": 8646.77930000471,
      "effective_prefill_tok_s": 1674.496971066013,
      "generation_ms": 6874.306700003217,
      "mean_inter_token_latency_ms": 229.1435566667739,
      "output_token_source": "usage",
      "output_tokens": 31,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 1772.4726000014925,
      "ttft_ms": 1772.473800003354,
      "visible_generation_ms": 6874.3055000013555
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 4.074827510332489,
      "e2e_ms": 13027.960500003246,
      "effective_prefill_tok_s": 401.9746707743686,
      "generation_ms": 5644.410700006119,
      "mean_inter_token_latency_ms": 245.40916086983128,
      "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": 7383.549799997127,
      "ttft_ms": 7383.5502000001725,
      "visible_generation_ms": 5644.410300003074
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 3.4394821562634807,
      "e2e_ms": 20529.72690000024,
      "effective_prefill_tok_s": 344.7423618834671,
      "generation_ms": 11920.399100003124,
      "mean_inter_token_latency_ms": 290.74144146349084,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 5,
      "time_to_first_output_ms": 8609.327799997118,
      "ttft_ms": 9861.01310000231,
      "visible_generation_ms": 10668.713799997931
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.504930816331262,
      "e2e_ms": 14851.066899995203,
      "effective_prefill_tok_s": 546.0138249671495,
      "generation_ms": 9415.307099996426,
      "mean_inter_token_latency_ms": 285.31233636352806,
      "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": 5435.759799998777,
      "ttft_ms": 5435.760299995309,
      "visible_generation_ms": 9415.306599999894
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 3.747471841875217,
      "e2e_ms": 19214.75510000164,
      "effective_prefill_tok_s": 444.77970097368024,
      "generation_ms": 12541.788700000325,
      "mean_inter_token_latency_ms": 266.8465680851133,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 6672.9664000013145,
      "ttft_ms": 6672.966999998607,
      "visible_generation_ms": 12541.788100003032
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 3.7213768966938847,
      "e2e_ms": 23598.64430000016,
      "effective_prefill_tok_s": 284.52352408946814,
      "generation_ms": 13167.169399996055,
      "mean_inter_token_latency_ms": 268.7177428570624,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 10431.474900004105,
      "ttft_ms": 10431.475400000636,
      "visible_generation_ms": 13167.168899999524
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 4.909133232515099,
      "e2e_ms": 20469.22970000014,
      "effective_prefill_tok_s": 240.88657036101,
      "generation_ms": 8148.077899997588,
      "mean_inter_token_latency_ms": 203.7019474999397,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 12321.151800002554,
      "ttft_ms": 12321.152299999085,
      "visible_generation_ms": 8148.077400001057
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 4.13471825470724,
      "e2e_ms": 24845.96590000001,
      "effective_prefill_tok_s": 208.94977800563044,
      "generation_ms": 10641.595699999016,
      "mean_inter_token_latency_ms": 241.85444772725035,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 14204.370200000994,
      "ttft_ms": 14204.370400002517,
      "visible_generation_ms": 10641.595499997493
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.5668012796062465,
      "e2e_ms": 24845.45029999572,
      "effective_prefill_tok_s": 184.50156516755672,
      "generation_ms": 8758.865899995726,
      "mean_inter_token_latency_ms": 218.97164749989315,
      "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": 16086.584399999992,
      "ttft_ms": 16086.584699995,
      "visible_generation_ms": 8758.865600000718
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.197912886708862,
      "e2e_ms": 22340.92409999721,
      "effective_prefill_tok_s": 284.5571178138074,
      "generation_ms": 11910.68069999892,
      "mean_inter_token_latency_ms": 238.2136139999784,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 12,
      "time_to_first_output_ms": 10430.24339999829,
      "ttft_ms": 12320.5089999974,
      "visible_generation_ms": 10020.41509999981
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.6789705158927175,
      "e2e_ms": 14203.404200001387,
      "effective_prefill_tok_s": 618.381489791656,
      "generation_ms": 9403.777999999875,
      "mean_inter_token_latency_ms": 213.72222727272444,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 4799.626200001512,
      "ttft_ms": 4799.626700005319,
      "visible_generation_ms": 9403.777499996067
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.363770369123167,
      "e2e_ms": 22340.11940000346,
      "effective_prefill_tok_s": 229.28524819956002,
      "generation_ms": 9395.544799997879,
      "mean_inter_token_latency_ms": 229.15962926824093,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 14,
      "time_to_first_output_ms": 12944.57460000558,
      "ttft_ms": 12944.575400004396,
      "visible_generation_ms": 9395.543999999063
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 4.439465928697213,
      "e2e_ms": 26094.33320000244,
      "effective_prefill_tok_s": 200.1116903711937,
      "generation_ms": 11262.61599999998,
      "mean_inter_token_latency_ms": 225.2523199999996,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 14831.717200002458,
      "ttft_ms": 14831.717700006266,
      "visible_generation_ms": 11262.615499996173
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.171075878437347,
      "e2e_ms": 18806.510999995226,
      "effective_prefill_tok_s": 393.71567175441515,
      "generation_ms": 11268.076000000292,
      "mean_inter_token_latency_ms": 239.74629787234664,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 16,
      "time_to_first_output_ms": 7538.434999994934,
      "ttft_ms": 8787.541999998211,
      "visible_generation_ms": 10018.968999997014
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 3.6929893766872244,
      "e2e_ms": 15664.008200001263,
      "effective_prefill_tok_s": 525.772730910479,
      "generation_ms": 10018.983599999046,
      "mean_inter_token_latency_ms": 270.78334054051476,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 17,
      "time_to_first_output_ms": 5645.024600002216,
      "ttft_ms": 7522.300299999188,
      "visible_generation_ms": 8141.707900002075
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.335400897430678,
      "e2e_ms": 14409.25900000002,
      "effective_prefill_tok_s": 525.8485850535287,
      "generation_ms": 8765.04869999917,
      "mean_inter_token_latency_ms": 230.65917631576764,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 18,
      "time_to_first_output_ms": 5644.21030000085,
      "ttft_ms": 5644.210899998143,
      "visible_generation_ms": 8765.048100001877
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 4.797452360900075,
      "e2e_ms": 16905.20039999683,
      "effective_prefill_tok_s": 430.1517850284198,
      "generation_ms": 10005.310399996233,
      "mean_inter_token_latency_ms": 208.44396666658818,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 19,
      "time_to_first_output_ms": 6899.890000000596,
      "ttft_ms": 6899.890399996366,
      "visible_generation_ms": 10005.310000000463
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 4.501248718994945,
      "e2e_ms": 13151.26710000186,
      "effective_prefill_tok_s": 473.79676414361546,
      "generation_ms": 6886.97780000075,
      "mean_inter_token_latency_ms": 222.1605741935726,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 20,
      "time_to_first_output_ms": 6264.28930000111,
      "ttft_ms": 6264.289900005679,
      "visible_generation_ms": 6886.977199996181
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 4.799040882884092,
      "e2e_ms": 13142.071399997803,
      "effective_prefill_tok_s": 430.7178259409809,
      "generation_ms": 6251.249100001587,
      "mean_inter_token_latency_ms": 208.3749700000529,
      "output_token_source": "usage",
      "output_tokens": 31,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 21,
      "time_to_first_output_ms": 6890.822299996216,
      "ttft_ms": 6890.822599998501,
      "visible_generation_ms": 6251.248799999303
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 5.128904450253344,
      "e2e_ms": 15074.274099999457,
      "effective_prefill_tok_s": 431.0575951625449,
      "generation_ms": 8188.883300004818,
      "mean_inter_token_latency_ms": 194.97341190487663,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 22,
      "time_to_first_output_ms": 6885.390799994639,
      "ttft_ms": 6885.391299998446,
      "visible_generation_ms": 8188.882800001011
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 5.936512457927807,
      "e2e_ms": 11500.188700003491,
      "effective_prefill_tok_s": 791.1435636341155,
      "generation_ms": 7748.657199997979,
      "mean_inter_token_latency_ms": 168.44906956517346,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 23,
      "time_to_first_output_ms": 3751.5315000055125,
      "ttft_ms": 5007.704499999818,
      "visible_generation_ms": 6492.484200003673
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 7.479488190529424,
      "e2e_ms": 10303.606699999364,
      "effective_prefill_tok_s": 790.9696635152638,
      "generation_ms": 6551.250399999844,
      "mean_inter_token_latency_ms": 133.69898775509887,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 24,
      "time_to_first_output_ms": 3752.3562999995193,
      "ttft_ms": 3752.3571999990963,
      "visible_generation_ms": 6551.249500000267
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 5.839936640308972,
      "e2e_ms": 9981.47450000397,
      "effective_prefill_tok_s": 591.7458000401813,
      "generation_ms": 4965.80730000278,
      "mean_inter_token_latency_ms": 171.23473448285446,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 25,
      "time_to_first_output_ms": 5015.6672000011895,
      "ttft_ms": 5015.667700004997,
      "visible_generation_ms": 4965.806799998973
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 12.012043165274203,
      "e2e_ms": 8676.334700001462,
      "effective_prefill_tok_s": 592.0199933608508,
      "generation_ms": 3662.9905000008876,
      "mean_inter_token_latency_ms": 83.24978409092927,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 26,
      "time_to_first_output_ms": 5013.344200000574,
      "ttft_ms": 5013.344699997106,
      "visible_generation_ms": 3662.9900000043563
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 8.570370265074041,
      "e2e_ms": 8537.75979999773,
      "effective_prefill_tok_s": 790.657918496112,
      "generation_ms": 4783.924000003026,
      "mean_inter_token_latency_ms": 116.6810731708055,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 27,
      "time_to_first_output_ms": 3753.835799994704,
      "ttft_ms": 3753.8361999977496,
      "visible_generation_ms": 4783.92359999998
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 13.004196229907237,
      "e2e_ms": 7233.70459999569,
      "effective_prefill_tok_s": 593.1663716562953,
      "generation_ms": 2230.0494000010076,
      "mean_inter_token_latency_ms": 76.89825517244854,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 28,
      "time_to_first_output_ms": 5003.655199994682,
      "ttft_ms": 5003.655599997728,
      "visible_generation_ms": 2230.048999997962
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 49.11521857279371,
      "e2e_ms": 6303.165099998296,
      "effective_prefill_tok_s": 544.783777924929,
      "generation_ms": 855.1320999977179,
      "mean_inter_token_latency_ms": 20.36028809518376,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 29,
      "time_to_first_output_ms": 5448.033000000578,
      "ttft_ms": 5448.033300002862,
      "visible_generation_ms": 855.1317999954335
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 15.601356171077379,
      "e2e_ms": 6127.055799996015,
      "effective_prefill_tok_s": 790.3144724668282,
      "generation_ms": 2371.5886999998474,
      "mean_inter_token_latency_ms": 64.09699189188777,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 30,
      "time_to_first_output_ms": 3755.4670999961672,
      "ttft_ms": 5447.740199997497,
      "visible_generation_ms": 679.3155999985174
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 51.406298862378144,
      "e2e_ms": 5091.097000004083,
      "effective_prefill_tok_s": 694.4190165995759,
      "generation_ms": 817.0205000060378,
      "mean_inter_token_latency_ms": 19.452869047762803,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 31,
      "time_to_first_output_ms": 4274.076499998046,
      "ttft_ms": 4274.076900001091,
      "visible_generation_ms": 817.020100002992
    }
  ],
  "requests": 32,
  "run_id": "benchmark-20260829T235833Z",
  "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": "2026-08-29T23:58:33Z",
  "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": 31288.693599999533
}
