{
  "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-b12x-fixed-mtp5-vision",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T22:30:19Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 499,
    "content_chunks_s": null,
    "decode_tok_s_p50": 14.871774168885684,
    "decode_tok_s_p95": 31.321120184518694,
    "e2e_p50_ms": 17578.132199996617,
    "e2e_p95_ms": 21098.337499999616,
    "effective_prefill_tok_s_p50": 197.24658389580793,
    "effective_prefill_tok_s_p95": 1303.8036450645504,
    "generation_p50_ms": 3227.5907000002917,
    "generation_p95_ms": 5409.300199993595,
    "mean_inter_token_latency_ms_p50": 66.37426097550792,
    "mean_inter_token_latency_ms_p95": 90.36656749995018,
    "output_token_sources": {
      "usage": 32
    },
    "output_tokens": 1560,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 32,
    "prompt_tokens": 94976,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 72,
    "throughput_tok_s": 40.764396081706884,
    "time_to_first_output_p50_ms": 14955.724100000225,
    "time_to_first_output_p95_ms": 17304.32490000385,
    "ttft_p50_ms": 15094.1995999965,
    "ttft_p95_ms": 18016.68570000038,
    "ttft_samples": 32
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-fixed-mtp5-vision",
  "request_timings": [
    {
      "content_chunks": 17,
      "decode_tok_s": 15.944874847171915,
      "e2e_ms": 14935.980400005064,
      "effective_prefill_tok_s": 248.87618524656412,
      "generation_ms": 3010.371700001997,
      "mean_inter_token_latency_ms": 62.716077083374934,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 11925.608700003067,
      "ttft_ms": 11925.609100006113,
      "visible_generation_ms": 3010.371299998951
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 17.950891781964586,
      "e2e_ms": 5769.3446000048425,
      "effective_prefill_tok_s": 976.4201122536133,
      "generation_ms": 2729.6694000033312,
      "mean_inter_token_latency_ms": 55.707538775578186,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 1,
      "time_to_first_output_ms": 3039.6752000015113,
      "ttft_ms": 4016.9220000025234,
      "visible_generation_ms": 1752.422600002319
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 13.042480106961847,
      "e2e_ms": 18293.33339999721,
      "effective_prefill_tok_s": 202.04599755492384,
      "generation_ms": 3603.609099998721,
      "mean_inter_token_latency_ms": 76.67253404252598,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 14689.724299998488,
      "ttft_ms": 14818.519000000379,
      "visible_generation_ms": 3474.8143999968306
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 12.559750220625311,
      "e2e_ms": 13644.08099999855,
      "effective_prefill_tok_s": 294.99442174293915,
      "generation_ms": 3582.873799998197,
      "mean_inter_token_latency_ms": 79.61941777773771,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 3,
      "time_to_first_output_ms": 10061.207200000354,
      "ttft_ms": 10187.227100002929,
      "visible_generation_ms": 3456.8538999956218
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 16.70611298149854,
      "e2e_ms": 10928.68499999895,
      "effective_prefill_tok_s": 335.98820237672004,
      "generation_ms": 2095.0415000042995,
      "mean_inter_token_latency_ms": 59.85832857155141,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 8833.64349999465,
      "ttft_ms": 8833.643899997696,
      "visible_generation_ms": 2095.0411000012537
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 12.720752044083188,
      "e2e_ms": 9001.2859999988,
      "effective_prefill_tok_s": 543.2157306050494,
      "generation_ms": 3537.526699998125,
      "mean_inter_token_latency_ms": 78.61170444440278,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 5,
      "time_to_first_output_ms": 5463.759300000675,
      "ttft_ms": 5591.908900001727,
      "visible_generation_ms": 3409.377099997073
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 15.066081117935152,
      "e2e_ms": 4353.855699999258,
      "effective_prefill_tok_s": 1818.058193784764,
      "generation_ms": 2721.344699995825,
      "mean_inter_token_latency_ms": 66.37426097550792,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 1632.5110000034329,
      "ttft_ms": 1632.5123999995412,
      "visible_generation_ms": 2721.3432999997167
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 22.31921159621869,
      "e2e_ms": 10162.73050000018,
      "effective_prefill_tok_s": 354.57635562270684,
      "generation_ms": 1792.1779999960563,
      "mean_inter_token_latency_ms": 44.80444999990141,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 8370.552500004123,
      "ttft_ms": 8370.553200002178,
      "visible_generation_ms": 1792.1772999980021
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 13.222916301261604,
      "e2e_ms": 5679.599099996267,
      "effective_prefill_tok_s": 1303.8036450645504,
      "generation_ms": 3403.1826999998884,
      "mean_inter_token_latency_ms": 75.62628222221974,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 8,
      "time_to_first_output_ms": 2276.4163999963785,
      "ttft_ms": 3608.9863999950467,
      "visible_generation_ms": 2070.61270000122
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 12.538013332982747,
      "e2e_ms": 19395.742900000187,
      "effective_prefill_tok_s": 186.8262915201347,
      "generation_ms": 3509.3278999993345,
      "mean_inter_token_latency_ms": 79.75745227271214,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 9,
      "time_to_first_output_ms": 15886.415000000852,
      "ttft_ms": 16012.6001000026,
      "visible_generation_ms": 3383.142799997586
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 11.646608187889923,
      "e2e_ms": 12430.763699994714,
      "effective_prefill_tok_s": 422.7038992653778,
      "generation_ms": 5409.300199993595,
      "mean_inter_token_latency_ms": 85.86190793640627,
      "output_token_source": "usage",
      "output_tokens": 64,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 10,
      "time_to_first_output_ms": 7021.4635000011185,
      "ttft_ms": 8353.738099998736,
      "visible_generation_ms": 4077.0255999959772
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 24.876492361516505,
      "e2e_ms": 5726.779600001464,
      "effective_prefill_tok_s": 757.5597275731442,
      "generation_ms": 1808.9366999993217,
      "mean_inter_token_latency_ms": 40.19859333331826,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 11,
      "time_to_first_output_ms": 3917.8429000021424,
      "ttft_ms": 4039.43380000419,
      "visible_generation_ms": 1687.3457999972743
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 16.15779835133193,
      "e2e_ms": 15951.191199994355,
      "effective_prefill_tok_s": 220.2498286088957,
      "generation_ms": 2475.5847999986145,
      "mean_inter_token_latency_ms": 61.88961999996536,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 12,
      "time_to_first_output_ms": 13475.60639999574,
      "ttft_ms": 13603.261199998087,
      "visible_generation_ms": 2347.9299999962677
    },
    {
      "content_chunks": 22,
      "decode_tok_s": 15.373096388457618,
      "e2e_ms": 21922.78290000104,
      "effective_prefill_tok_s": 171.517815179218,
      "generation_ms": 4618.45799999719,
      "mean_inter_token_latency_ms": 65.04870422531253,
      "output_token_source": "usage",
      "output_tokens": 72,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 13,
      "time_to_first_output_ms": 17304.32490000385,
      "ttft_ms": 18016.68570000038,
      "visible_generation_ms": 3906.0972000006586
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 15.128466558708794,
      "e2e_ms": 15950.43280000391,
      "effective_prefill_tok_s": 243.624129568042,
      "generation_ms": 3767.7315000037197,
      "mean_inter_token_latency_ms": 66.1005526316442,
      "output_token_source": "usage",
      "output_tokens": 58,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 14,
      "time_to_first_output_ms": 12182.70130000019,
      "ttft_ms": 12309.158400006709,
      "visible_generation_ms": 3641.274399997201
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 16.40330073779189,
      "e2e_ms": 20683.11890000041,
      "effective_prefill_tok_s": 165.44249071305276,
      "generation_ms": 2743.3502999992925,
      "mean_inter_token_latency_ms": 60.96333999998428,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 15,
      "time_to_first_output_ms": 17939.768600001116,
      "ttft_ms": 18078.89989999967,
      "visible_generation_ms": 2604.2190000007395
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 15.102259785580824,
      "e2e_ms": 18712.713599998096,
      "effective_prefill_tok_s": 198.6820046318152,
      "generation_ms": 3774.269599998661,
      "mean_inter_token_latency_ms": 66.21525614032738,
      "output_token_source": "usage",
      "output_tokens": 58,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 16,
      "time_to_first_output_ms": 14938.443999999436,
      "ttft_ms": 15069.060699999682,
      "visible_generation_ms": 3643.652899998415
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 15.69623750046719,
      "e2e_ms": 18719.579700002214,
      "effective_prefill_tok_s": 200.95346596827164,
      "generation_ms": 3949.9912000028417,
      "mean_inter_token_latency_ms": 63.7095354839168,
      "output_token_source": "usage",
      "output_tokens": 63,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 17,
      "time_to_first_output_ms": 14769.588499999372,
      "ttft_ms": 14896.307700000762,
      "visible_generation_ms": 3823.2720000014524
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 14.028893273305874,
      "e2e_ms": 19836.383900001238,
      "effective_prefill_tok_s": 184.82462195255255,
      "generation_ms": 3777.9174000024796,
      "mean_inter_token_latency_ms": 71.28146037740528,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 18,
      "time_to_first_output_ms": 16058.466499998758,
      "ttft_ms": 16190.552500003832,
      "visible_generation_ms": 3645.8313999974052
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 14.871774168885684,
      "e2e_ms": 20468.2285999952,
      "effective_prefill_tok_s": 172.15140282024464,
      "generation_ms": 3227.5907000002917,
      "mean_inter_token_latency_ms": 67.24147291667275,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 19,
      "time_to_first_output_ms": 17240.63789999491,
      "ttft_ms": 17367.242599997553,
      "visible_generation_ms": 3100.985999997647
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 12.069018471665375,
      "e2e_ms": 17644.598099999712,
      "effective_prefill_tok_s": 194.72827647140195,
      "generation_ms": 2402.8466000017943,
      "mean_inter_token_latency_ms": 82.8567793104067,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 20,
      "time_to_first_output_ms": 15241.751499997918,
      "ttft_ms": 15241.751900000963,
      "visible_generation_ms": 2402.8461999987485
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 14.108128514425259,
      "e2e_ms": 18906.944399997883,
      "effective_prefill_tok_s": 194.0883221274829,
      "generation_ms": 3614.937300000747,
      "mean_inter_token_latency_ms": 70.88112352942642,
      "output_token_source": "usage",
      "output_tokens": 52,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 21,
      "time_to_first_output_ms": 15292.007099997136,
      "ttft_ms": 15419.538800000737,
      "visible_generation_ms": 3487.405599997146
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 11.066039439868636,
      "e2e_ms": 19337.042099999962,
      "effective_prefill_tok_s": 188.77549793764874,
      "generation_ms": 3614.662699998007,
      "mean_inter_token_latency_ms": 90.36656749995018,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 22,
      "time_to_first_output_ms": 15722.379400001955,
      "ttft_ms": 17034.028499998385,
      "visible_generation_ms": 2303.0136000015773
    },
    {
      "content_chunks": 22,
      "decode_tok_s": 13.533756627045777,
      "e2e_ms": 20015.3674999965,
      "effective_prefill_tok_s": 192.29984262254774,
      "generation_ms": 4581.137499997567,
      "mean_inter_token_latency_ms": 73.88931451608978,
      "output_token_source": "usage",
      "output_tokens": 63,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 23,
      "time_to_first_output_ms": 15434.229999998934,
      "ttft_ms": 16151.354800000263,
      "visible_generation_ms": 3864.0126999962376
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 19.334892807527464,
      "e2e_ms": 17852.04220000014,
      "effective_prefill_tok_s": 198.45244403779523,
      "generation_ms": 2896.3180999999167,
      "mean_inter_token_latency_ms": 51.71996607142709,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 24,
      "time_to_first_output_ms": 14955.724100000225,
      "ttft_ms": 15094.1995999965,
      "visible_generation_ms": 2757.8426000036416
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 13.322725797967374,
      "e2e_ms": 18789.952899998752,
      "effective_prefill_tok_s": 194.4680611533237,
      "generation_ms": 3527.806599995529,
      "mean_inter_token_latency_ms": 75.05971489352189,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 25,
      "time_to_first_output_ms": 15262.146300003224,
      "ttft_ms": 15388.605400003144,
      "visible_generation_ms": 3401.347499995609
    },
    {
      "content_chunks": 23,
      "decode_tok_s": 10.62167971858971,
      "e2e_ms": 21098.337499999616,
      "effective_prefill_tok_s": 194.47987357084725,
      "generation_ms": 5837.118200004625,
      "mean_inter_token_latency_ms": 94.14706774201008,
      "output_token_source": "usage",
      "output_tokens": 63,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 26,
      "time_to_first_output_ms": 15261.219299994991,
      "ttft_ms": 15388.576399993326,
      "visible_generation_ms": 5709.76110000629
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 13.794377119756314,
      "e2e_ms": 19026.21270000236,
      "effective_prefill_tok_s": 179.21146628774838,
      "generation_ms": 2464.7724000024027,
      "mean_inter_token_latency_ms": 72.49330588242361,
      "output_token_source": "usage",
      "output_tokens": 35,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 27,
      "time_to_first_output_ms": 16561.44029999996,
      "ttft_ms": 16561.440600002243,
      "visible_generation_ms": 2464.7721000001184
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 22.699734789067502,
      "e2e_ms": 16783.334899999318,
      "effective_prefill_tok_s": 193.05789314465014,
      "generation_ms": 1409.708099999989,
      "mean_inter_token_latency_ms": 44.053378124999654,
      "output_token_source": "usage",
      "output_tokens": 33,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 28,
      "time_to_first_output_ms": 15373.626799999329,
      "ttft_ms": 15373.627200002375,
      "visible_generation_ms": 1409.7076999969431
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 13.000825417912521,
      "e2e_ms": 17578.132199996617,
      "effective_prefill_tok_s": 193.3864870775796,
      "generation_ms": 2230.6276000017533,
      "mean_inter_token_latency_ms": 76.91819310350874,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 29,
      "time_to_first_output_ms": 15347.504599994863,
      "ttft_ms": 15347.504899997148,
      "visible_generation_ms": 2230.627299999469
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 31.321120184518694,
      "e2e_ms": 17149.6019000042,
      "effective_prefill_tok_s": 189.66047515951712,
      "generation_ms": 1500.584900000831,
      "mean_inter_token_latency_ms": 31.927338297890024,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 30,
      "time_to_first_output_ms": 15649.017000003369,
      "ttft_ms": 16365.178100000776,
      "visible_generation_ms": 784.4238000034238
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 49.299949079165444,
      "e2e_ms": 16243.911199999275,
      "effective_prefill_tok_s": 197.24658389580793,
      "generation_ms": 1196.7557999960263,
      "mean_inter_token_latency_ms": 20.28399661010214,
      "output_token_source": "usage",
      "output_tokens": 60,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 31,
      "time_to_first_output_ms": 15047.155400003248,
      "ttft_ms": 15177.469599999313,
      "visible_generation_ms": 1066.441599999962
    }
  ],
  "requests": 32,
  "run_id": "benchmark-20260829T222941Z",
  "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": "2026-08-29T22:29:41Z",
  "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": 38268.68909999757
}
