{
  "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": "2026-08-29T23:46:42Z",
  "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": 460,
    "content_chunks_s": null,
    "decode_tok_s_p50": 4.457420363942551,
    "decode_tok_s_p95": 5.638906802857773,
    "e2e_p50_ms": 17465.079699999478,
    "e2e_p95_ms": 23067.018999994616,
    "effective_prefill_tok_s_p50": 395.90068319138237,
    "effective_prefill_tok_s_p95": 1023.56637424434,
    "generation_p50_ms": 8720.539099995221,
    "generation_p95_ms": 13696.722699998645,
    "mean_inter_token_latency_ms_p50": 221.5272187499977,
    "mean_inter_token_latency_ms_p95": 315.61919523803573,
    "output_token_sources": {
      "usage": 32
    },
    "output_tokens": 1318,
    "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": 35.01719567482515,
    "time_to_first_output_p50_ms": 7494.369000000006,
    "time_to_first_output_p95_ms": 14808.029400002852,
    "ttft_p50_ms": 7496.830000003683,
    "ttft_p95_ms": 16071.672400001262,
    "ttft_samples": 32
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 10,
      "decode_tok_s": 4.677693549592492,
      "e2e_ms": 22935.47170000238,
      "effective_prefill_tok_s": 192.06437998017458,
      "generation_ms": 7482.320000002801,
      "mean_inter_token_latency_ms": 213.78057142865146,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 0,
      "time_to_first_output_ms": 15453.15169999958,
      "ttft_ms": 16689.02109999908,
      "visible_generation_ms": 6246.4506000033
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 3.686138216335387,
      "e2e_ms": 18563.058300002012,
      "effective_prefill_tok_s": 240.841336689201,
      "generation_ms": 6239.592400001129,
      "mean_inter_token_latency_ms": 271.28662608700563,
      "output_token_source": "usage",
      "output_tokens": 24,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 12323.465900000883,
      "ttft_ms": 12323.466499998176,
      "visible_generation_ms": 6239.591800003836
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 4.4374050218122525,
      "e2e_ms": 21059.468200000993,
      "effective_prefill_tok_s": 229.24905063050952,
      "generation_ms": 8112.849700002698,
      "mean_inter_token_latency_ms": 225.35693611118606,
      "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": 12946.618499998294,
      "ttft_ms": 14192.652400000952,
      "visible_generation_ms": 6866.81580000004
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 4.029287926912431,
      "e2e_ms": 10471.748099997058,
      "effective_prefill_tok_s": 980.7472831685765,
      "generation_ms": 7445.484299998498,
      "mean_inter_token_latency_ms": 248.1828099999499,
      "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": 3026.26379999856,
      "ttft_ms": 3026.2643000023672,
      "visible_generation_ms": 7445.48379999469
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 5.0515396267271075,
      "e2e_ms": 19192.374400001427,
      "effective_prefill_tok_s": 267.9658798539851,
      "generation_ms": 8116.3374000025215,
      "mean_inter_token_latency_ms": 197.9594487805493,
      "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": 11076.036999998905,
      "ttft_ms": 11076.037200000428,
      "visible_generation_ms": 8116.337200000999
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.514117974498384,
      "e2e_ms": 25441.33590000274,
      "effective_prefill_tok_s": 200.43180087145345,
      "generation_ms": 10633.30649999989,
      "mean_inter_token_latency_ms": 221.5272187499977,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 5,
      "time_to_first_output_ms": 14808.029400002852,
      "ttft_ms": 16071.672400001262,
      "visible_generation_ms": 9369.66350000148
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 4.805588112608381,
      "e2e_ms": 21030.003799998667,
      "effective_prefill_tok_s": 209.56014659689436,
      "generation_ms": 6867.005499996594,
      "mean_inter_token_latency_ms": 208.09107575747254,
      "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": 14162.998300002073,
      "ttft_ms": 14162.998599997081,
      "visible_generation_ms": 6867.005200001586
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 3.5846466026100443,
      "e2e_ms": 17250.10889999976,
      "effective_prefill_tok_s": 324.015677380535,
      "generation_ms": 8090.058299996599,
      "mean_inter_token_latency_ms": 278.9675275860896,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 9160.05060000316,
      "ttft_ms": 9160.05099999893,
      "visible_generation_ms": 8090.057900000829
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 4.5989447675220765,
      "e2e_ms": 10370.465100000729,
      "effective_prefill_tok_s": 1774.251910417767,
      "generation_ms": 8697.647400003916,
      "mean_inter_token_latency_ms": 217.4411850000979,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 1672.8176999968127,
      "ttft_ms": 1672.819100000197,
      "visible_generation_ms": 8697.646000000532
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.586872387284167,
      "e2e_ms": 19067.927799995232,
      "effective_prefill_tok_s": 286.8356535209697,
      "generation_ms": 8720.539099995221,
      "mean_inter_token_latency_ms": 218.01347749988054,
      "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": 10347.38870000001,
      "ttft_ms": 10347.388899994257,
      "visible_generation_ms": 8720.538900000975
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 4.420837795782375,
      "e2e_ms": 18435.81450000056,
      "effective_prefill_tok_s": 349.8782893909389,
      "generation_ms": 9952.864600003522,
      "mean_inter_token_latency_ms": 226.20146818189824,
      "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": 8482.949899997038,
      "ttft_ms": 8482.950300000084,
      "visible_generation_ms": 9952.864200000477
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.60097836814605,
      "e2e_ms": 14066.37480000063,
      "effective_prefill_tok_s": 552.435624078711,
      "generation_ms": 8693.80309999542,
      "mean_inter_token_latency_ms": 217.34507749988552,
      "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": 5372.57170000521,
      "ttft_ms": 5372.57210000098,
      "visible_generation_ms": 8693.802699999651
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.457420363942551,
      "e2e_ms": 13443.880699996953,
      "effective_prefill_tok_s": 1023.56637424434,
      "generation_ms": 10544.215299996722,
      "mean_inter_token_latency_ms": 224.345006382909,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 12,
      "time_to_first_output_ms": 2899.665400000231,
      "ttft_ms": 4130.334200002835,
      "visible_generation_ms": 9313.546499994118
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 5.638906802857773,
      "e2e_ms": 14066.092200002458,
      "effective_prefill_tok_s": 448.4847019572033,
      "generation_ms": 7448.2522000034805,
      "mean_inter_token_latency_ms": 177.33933809532095,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 6617.839999998978,
      "ttft_ms": 6617.840400002024,
      "visible_generation_ms": 7448.251800000435
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.726503062417946,
      "e2e_ms": 16561.74219999957,
      "effective_prefill_tok_s": 448.4863623047943,
      "generation_ms": 9943.926699997974,
      "mean_inter_token_latency_ms": 211.57290851059517,
      "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": 6617.815500001598,
      "ttft_ms": 7849.827700003516,
      "visible_generation_ms": 8711.914499996055
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 5.006146740755795,
      "e2e_ms": 10948.934899999585,
      "effective_prefill_tok_s": 623.9819953441767,
      "generation_ms": 6192.387399998552,
      "mean_inter_token_latency_ms": 199.7544322580178,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 4756.547500001034,
      "ttft_ms": 4756.547899996804,
      "visible_generation_ms": 6192.387000002782
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.598974244783855,
      "e2e_ms": 16218.928100002813,
      "effective_prefill_tok_s": 477.4238901787469,
      "generation_ms": 10002.230400001281,
      "mean_inter_token_latency_ms": 217.43979130437566,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 16,
      "time_to_first_output_ms": 6216.697700001532,
      "ttft_ms": 7463.266400001885,
      "visible_generation_ms": 8755.661700000928
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 4.799008236955116,
      "e2e_ms": 17465.079699999478,
      "effective_prefill_tok_s": 397.6945987140759,
      "generation_ms": 10002.066600005492,
      "mean_inter_token_latency_ms": 208.37638750011442,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 17,
      "time_to_first_output_ms": 7463.0130999939865,
      "ttft_ms": 7463.013499997032,
      "visible_generation_ms": 10002.066200002446
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 4.1458206090793155,
      "e2e_ms": 18728.64639999898,
      "effective_prefill_tok_s": 365.7177330652007,
      "generation_ms": 10613.097900000867,
      "mean_inter_token_latency_ms": 241.20677045456515,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 18,
      "time_to_first_output_ms": 8115.548499998113,
      "ttft_ms": 8115.549699999974,
      "visible_generation_ms": 10613.096699999005
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 4.802430942520262,
      "e2e_ms": 16234.068799996749,
      "effective_prefill_tok_s": 475.70729685148166,
      "generation_ms": 9994.938099996943,
      "mean_inter_token_latency_ms": 208.22787708326965,
      "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": 6239.130699999805,
      "ttft_ms": 6239.131299997098,
      "visible_generation_ms": 9994.93749999965
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.911954079855784,
      "e2e_ms": 16236.958000001323,
      "effective_prefill_tok_s": 396.6427293401512,
      "generation_ms": 8754.15349999821,
      "mean_inter_token_latency_ms": 203.58496511623744,
      "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": 7482.8045000031125,
      "ttft_ms": 7482.804900006158,
      "visible_generation_ms": 8754.153099995165
    },
    {
      "content_chunks": 23,
      "decode_tok_s": 3.802528921768418,
      "e2e_ms": 23067.018999994616,
      "effective_prefill_tok_s": 475.93735766392047,
      "generation_ms": 16830.90419999644,
      "mean_inter_token_latency_ms": 262.98287812494436,
      "output_token_source": "usage",
      "output_tokens": 65,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 21,
      "time_to_first_output_ms": 6236.114799998177,
      "ttft_ms": 7482.396399995196,
      "visible_generation_ms": 15584.62259999942
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 2.993416812038113,
      "e2e_ms": 19942.46810000186,
      "effective_prefill_tok_s": 475.20348812144533,
      "generation_ms": 13696.722699998645,
      "mean_inter_token_latency_ms": 334.0664073170401,
      "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": 6245.745400003216,
      "ttft_ms": 6245.745799998986,
      "visible_generation_ms": 13696.722300002875
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.1683751022995086,
      "e2e_ms": 19499.653099999705,
      "effective_prefill_tok_s": 475.36320479605473,
      "generation_ms": 13256.0061999975,
      "mean_inter_token_latency_ms": 315.61919523803573,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 23,
      "time_to_first_output_ms": 6243.646900002204,
      "ttft_ms": 6243.647299997974,
      "visible_generation_ms": 13256.00580000173
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 3.9125316344161782,
      "e2e_ms": 18783.88850000192,
      "effective_prefill_tok_s": 474.11927751532437,
      "generation_ms": 12523.86039999692,
      "mean_inter_token_latency_ms": 255.58898775503917,
      "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": 6260.028100005002,
      "ttft_ms": 6260.029299999587,
      "visible_generation_ms": 12523.859200002335
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 4.078070077095083,
      "e2e_ms": 18783.197600001586,
      "effective_prefill_tok_s": 395.55653538421234,
      "generation_ms": 11279.84540000034,
      "mean_inter_token_latency_ms": 245.21403043479,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 25,
      "time_to_first_output_ms": 7503.352200001245,
      "ttft_ms": 7503.352599997015,
      "visible_generation_ms": 11279.84500000457
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.6721687120210253,
      "e2e_ms": 18275.168499996653,
      "effective_prefill_tok_s": 339.4321327867019,
      "generation_ms": 9531.152499999735,
      "mean_inter_token_latency_ms": 272.3186428571353,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 26,
      "time_to_first_output_ms": 8744.015999996918,
      "ttft_ms": 8744.016299999203,
      "visible_generation_ms": 9531.15219999745
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.7843803862379426,
      "e2e_ms": 18067.522900004406,
      "effective_prefill_tok_s": 395.8515134829024,
      "generation_ms": 10569.761999999173,
      "mean_inter_token_latency_ms": 264.2440499999793,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 27,
      "time_to_first_output_ms": 7497.7609000052325,
      "ttft_ms": 7497.761400001764,
      "visible_generation_ms": 10569.761500002642
    },
    {
      "content_chunks": 8,
      "decode_tok_s": 3.4347717363796404,
      "e2e_ms": 15939.89440000587,
      "effective_prefill_tok_s": 395.90068319138237,
      "generation_ms": 8443.06470000447,
      "mean_inter_token_latency_ms": 291.14016206911964,
      "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": 7496.829700001399,
      "ttft_ms": 7496.830000003683,
      "visible_generation_ms": 8443.064400002186
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 5.490293556802041,
      "e2e_ms": 16576.87500000611,
      "effective_prefill_tok_s": 339.39893820261176,
      "generation_ms": 7832.003800001985,
      "mean_inter_token_latency_ms": 182.1396232558601,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 29,
      "time_to_first_output_ms": 8744.871200004127,
      "ttft_ms": 8744.871700000658,
      "visible_generation_ms": 7832.003300005454
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 4.40432229960861,
      "e2e_ms": 14532.909300003666,
      "effective_prefill_tok_s": 396.0306731627436,
      "generation_ms": 7038.54030000366,
      "mean_inter_token_latency_ms": 227.04968709689226,
      "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": 7494.369000000006,
      "ttft_ms": 7494.369400003052,
      "visible_generation_ms": 7038.539900000615
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 37.19520330653822,
      "e2e_ms": 12116.693200005102,
      "effective_prefill_tok_s": 263.0447228140455,
      "generation_ms": 833.4408000009716,
      "mean_inter_token_latency_ms": 26.885187096805538,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 31,
      "time_to_first_output_ms": 11283.25240000413,
      "ttft_ms": 11283.252699999139,
      "visible_generation_ms": 833.4405000059633
    }
  ],
  "requests": 32,
  "run_id": "benchmark-20260829T234605Z",
  "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": "2026-08-29T23:46:05Z",
  "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": 37638.650799999596
}
