{
  "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": "2026-08-29T23:58:23Z",
  "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": 501,
    "content_chunks_s": null,
    "decode_tok_s_p50": 3.9949461711544756,
    "decode_tok_s_p95": 7.139206560577152,
    "e2e_p50_ms": 17167.085699999006,
    "e2e_p95_ms": 23919.506900005217,
    "effective_prefill_tok_s_p50": 429.93254086132055,
    "effective_prefill_tok_s_p95": 1056.550400673946,
    "generation_p50_ms": 9489.403200001107,
    "generation_p95_ms": 15146.555699997407,
    "mean_inter_token_latency_ms_p50": 250.15430833324595,
    "mean_inter_token_latency_ms_p95": 365.10822580631583,
    "output_token_sources": {
      "usage": 32
    },
    "output_tokens": 1371,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 32,
    "prompt_tokens": 94976,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 33,
    "throughput_tok_s": 35.77153059987306,
    "time_to_first_output_p50_ms": 6774.013000002014,
    "time_to_first_output_p95_ms": 15749.438400001964,
    "ttft_p50_ms": 7030.505000002449,
    "ttft_p95_ms": 15749.439100000018,
    "ttft_samples": 32
  },
  "model": "glm53-flash-exl3-k3-dflash2-k3-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 3.902979899133373,
      "e2e_ms": 12032.863500004169,
      "effective_prefill_tok_s": 1056.550400673946,
      "generation_ms": 9223.721599999408,
      "mean_inter_token_latency_ms": 256.21448888887244,
      "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": 2809.141900004761,
      "ttft_ms": 4841.239200002747,
      "visible_generation_ms": 7191.624300001422
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 4.332319298986051,
      "e2e_ms": 13907.23170000274,
      "effective_prefill_tok_s": 704.5437421728986,
      "generation_ms": 9694.576300003973,
      "mean_inter_token_latency_ms": 230.82324523818983,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 4212.655399998766,
      "ttft_ms": 4212.659199998598,
      "visible_generation_ms": 9694.572500004142
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 4.067548972272827,
      "e2e_ms": 22776.144600000407,
      "effective_prefill_tok_s": 173.3479440107347,
      "generation_ms": 5654.510899999877,
      "mean_inter_token_latency_ms": 245.84829999999468,
      "output_token_source": "usage",
      "output_tokens": 24,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 17121.63370000053,
      "ttft_ms": 17121.634200004337,
      "visible_generation_ms": 5654.51039999607
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 3.9975325896359877,
      "e2e_ms": 16499.70839999878,
      "effective_prefill_tok_s": 396.04207189079483,
      "generation_ms": 9005.555099996855,
      "mean_inter_token_latency_ms": 250.15430833324595,
      "output_token_source": "usage",
      "output_tokens": 37,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 3,
      "time_to_first_output_ms": 7494.153300001926,
      "ttft_ms": 9506.204100005561,
      "visible_generation_ms": 6993.504299993219
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 3.864138834313883,
      "e2e_ms": 26586.353599996073,
      "effective_prefill_tok_s": 213.43842558153563,
      "generation_ms": 12680.70379999699,
      "mean_inter_token_latency_ms": 258.7898734693263,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 13905.649799999082,
      "ttft_ms": 13905.65069999866,
      "visible_generation_ms": 12680.702899997414
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 2.7389139146114014,
      "e2e_ms": 12602.67659999954,
      "effective_prefill_tok_s": 2310.9476629462074,
      "generation_ms": 11318.35499999579,
      "mean_inter_token_latency_ms": 365.10822580631583,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 5,
      "time_to_first_output_ms": 1284.32160000375,
      "ttft_ms": 2757.7161000008346,
      "visible_generation_ms": 9844.960499998706
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 4.247269630581845,
      "e2e_ms": 17059.293600003002,
      "effective_prefill_tok_s": 336.5573364090033,
      "generation_ms": 8240.588200002094,
      "mean_inter_token_latency_ms": 235.44537714291695,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 6,
      "time_to_first_output_ms": 8818.705400000908,
      "ttft_ms": 10695.75320000149,
      "visible_generation_ms": 6363.540400001511
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 3.9949461711544756,
      "e2e_ms": 15785.398499996518,
      "effective_prefill_tok_s": 438.14501094094703,
      "generation_ms": 9011.385499994503,
      "mean_inter_token_latency_ms": 250.31626388873622,
      "output_token_source": "usage",
      "output_tokens": 37,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 6774.013000002014,
      "ttft_ms": 8792.584499999066,
      "visible_generation_ms": 6992.813999997452
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.8815604165661006,
      "e2e_ms": 15128.44960000075,
      "effective_prefill_tok_s": 485.6452079661024,
      "generation_ms": 9016.99219999864,
      "mean_inter_token_latency_ms": 257.6283485713897,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 6111.457400002109,
      "ttft_ms": 6111.4581000001635,
      "visible_generation_ms": 9016.991500000586
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 4.171852918777793,
      "e2e_ms": 19504.095800002688,
      "effective_prefill_tok_s": 279.07497177975273,
      "generation_ms": 8868.960799998604,
      "mean_inter_token_latency_ms": 239.7016432432055,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 9,
      "time_to_first_output_ms": 10635.135000004084,
      "ttft_ms": 12530.288499998278,
      "visible_generation_ms": 6973.80730000441
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 4.727149610669337,
      "e2e_ms": 11909.768700003042,
      "effective_prefill_tok_s": 554.5688249807214,
      "generation_ms": 6557.863100002578,
      "mean_inter_token_latency_ms": 211.5439709678251,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 5351.905600000464,
      "ttft_ms": 5351.906500000041,
      "visible_generation_ms": 6557.862200003001
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.109847498101403,
      "e2e_ms": 22652.12690000044,
      "effective_prefill_tok_s": 225.4852466439108,
      "generation_ms": 9489.403200001107,
      "mean_inter_token_latency_ms": 243.31803076925914,
      "output_token_source": "usage",
      "output_tokens": 40,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 11,
      "time_to_first_output_ms": 13162.723699999333,
      "ttft_ms": 13162.724299996626,
      "visible_generation_ms": 9489.402600003814
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 5.018317777872202,
      "e2e_ms": 23919.506900005217,
      "effective_prefill_tok_s": 188.45116407449996,
      "generation_ms": 8170.068500003254,
      "mean_inter_token_latency_ms": 199.26996341471352,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 15749.438400001964,
      "ttft_ms": 15749.439100000018,
      "visible_generation_ms": 8170.067800005199
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 4.547771273163992,
      "e2e_ms": 23918.670599996403,
      "effective_prefill_tok_s": 196.25536433261695,
      "generation_ms": 8795.517099999415,
      "mean_inter_token_latency_ms": 219.88792749998538,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 15123.153499996988,
      "ttft_ms": 15123.154100001557,
      "visible_generation_ms": 8795.516499994847
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 4.556242273218497,
      "e2e_ms": 20759.164499999315,
      "effective_prefill_tok_s": 296.6609223929238,
      "generation_ms": 10754.476399997657,
      "mean_inter_token_latency_ms": 219.47911020403382,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 14,
      "time_to_first_output_ms": 10004.688100001658,
      "ttft_ms": 10004.688499997428,
      "visible_generation_ms": 10754.476000001887
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 3.8687855993659213,
      "e2e_ms": 23278.96490000421,
      "effective_prefill_tok_s": 249.28844024807222,
      "generation_ms": 11373.077900003409,
      "mean_inter_token_latency_ms": 258.47904318189563,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 11905.887000000803,
      "ttft_ms": 11905.887599998096,
      "visible_generation_ms": 11373.077300006116
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 3.972073743481612,
      "e2e_ms": 18432.089100002486,
      "effective_prefill_tok_s": 583.2263855705179,
      "generation_ms": 13343.156100003398,
      "mean_inter_token_latency_ms": 251.75766226421507,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 16,
      "time_to_first_output_ms": 5088.932999999088,
      "ttft_ms": 6341.6700000016135,
      "visible_generation_ms": 12090.419100000872
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 3.887137958381114,
      "e2e_ms": 18431.92079999426,
      "effective_prefill_tok_s": 468.0824984681424,
      "generation_ms": 12091.158199997153,
      "mean_inter_token_latency_ms": 257.2586851063224,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 17,
      "time_to_first_output_ms": 6340.762599997106,
      "ttft_ms": 7593.101399994339,
      "visible_generation_ms": 10838.81939999992
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 4.613028493338357,
      "e2e_ms": 17811.71830000676,
      "effective_prefill_tok_s": 425.6508076075834,
      "generation_ms": 10838.866500002041,
      "mean_inter_token_latency_ms": 216.77733000004082,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 18,
      "time_to_first_output_ms": 6972.851800004719,
      "ttft_ms": 6972.852200000489,
      "visible_generation_ms": 10838.866100006271
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 4.90635042870304,
      "e2e_ms": 17167.085699999006,
      "effective_prefill_tok_s": 425.445816830228,
      "generation_ms": 10190.874199994141,
      "mean_inter_token_latency_ms": 203.81748399988282,
      "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": 6976.211500004865,
      "ttft_ms": 6976.2120000013965,
      "visible_generation_ms": 10190.87369999761
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 3.4398517006304288,
      "e2e_ms": 22206.33739999903,
      "effective_prefill_tok_s": 526.6281130412906,
      "generation_ms": 16570.48180000129,
      "mean_inter_token_latency_ms": 290.71020701756646,
      "output_token_source": "usage",
      "output_tokens": 58,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 20,
      "time_to_first_output_ms": 5635.85559999774,
      "ttft_ms": 7527.530200000911,
      "visible_generation_ms": 14678.807199998118
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.2350589117767803,
      "e2e_ms": 21413.68999999395,
      "effective_prefill_tok_s": 473.58168150340066,
      "generation_ms": 15146.555699997407,
      "mean_inter_token_latency_ms": 309.1133816326001,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 21,
      "time_to_first_output_ms": 6267.134299996542,
      "ttft_ms": 8169.702399995003,
      "visible_generation_ms": 13243.987599998945
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 2.4791210430920514,
      "e2e_ms": 20617.946699996537,
      "effective_prefill_tok_s": 429.93254086132055,
      "generation_ms": 13714.538100000937,
      "mean_inter_token_latency_ms": 403.3687676470864,
      "output_token_source": "usage",
      "output_tokens": 35,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 22,
      "time_to_first_output_ms": 6903.4085999956005,
      "ttft_ms": 6903.409099999408,
      "visible_generation_ms": 13714.53759999713
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 3.6299130048380817,
      "e2e_ms": 20419.91930000222,
      "effective_prefill_tok_s": 428.84142363089165,
      "generation_ms": 13498.946100000467,
      "mean_inter_token_latency_ms": 275.4886959183769,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 23,
      "time_to_first_output_ms": 6920.973200001754,
      "ttft_ms": 8194.35810000141,
      "visible_generation_ms": 12225.561200000811
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 3.9545669966632504,
      "e2e_ms": 18079.55740000034,
      "effective_prefill_tok_s": 521.7250402514949,
      "generation_ms": 12390.737100002298,
      "mean_inter_token_latency_ms": 252.8721857143326,
      "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": 5688.820299998042,
      "ttft_ms": 5688.820699993812,
      "visible_generation_ms": 12390.736700006528
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 2.7810315847329043,
      "e2e_ms": 16489.968499998213,
      "effective_prefill_tok_s": 520.4639388894901,
      "generation_ms": 10787.36399999616,
      "mean_inter_token_latency_ms": 359.578799999872,
      "output_token_source": "usage",
      "output_tokens": 31,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 25,
      "time_to_first_output_ms": 5702.604500002053,
      "ttft_ms": 5702.604999998584,
      "visible_generation_ms": 10787.363499999628
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.8603190269928978,
      "e2e_ms": 14820.978300005663,
      "effective_prefill_tok_s": 665.5991365507625,
      "generation_ms": 10361.837900003593,
      "mean_inter_token_latency_ms": 259.0459475000898,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 26,
      "time_to_first_output_ms": 4459.14040000207,
      "ttft_ms": 4459.140700004355,
      "visible_generation_ms": 10361.837600001309
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 4.595611957938191,
      "e2e_ms": 15125.854000005347,
      "effective_prefill_tok_s": 514.4647894352049,
      "generation_ms": 9356.75170000468,
      "mean_inter_token_latency_ms": 217.59887674429487,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 27,
      "time_to_first_output_ms": 5769.102300000668,
      "ttft_ms": 5769.102900005237,
      "visible_generation_ms": 9356.75110000011
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 3.657595107150434,
      "e2e_ms": 14253.085199998168,
      "effective_prefill_tok_s": 513.7107103902049,
      "generation_ms": 8475.514399993699,
      "mean_inter_token_latency_ms": 273.4036903223774,
      "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": 5777.570800004469,
      "ttft_ms": 5777.571200000239,
      "visible_generation_ms": 8475.51399999793
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 7.139206560577152,
      "e2e_ms": 14034.083999998984,
      "effective_prefill_tok_s": 422.1603098019809,
      "generation_ms": 7003.57939999958,
      "mean_inter_token_latency_ms": 140.0715879999916,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 29,
      "time_to_first_output_ms": 7030.504599999404,
      "ttft_ms": 7030.505000002449,
      "visible_generation_ms": 7003.578999996535
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 6.43617895456474,
      "e2e_ms": 14034.058400000504,
      "effective_prefill_tok_s": 462.244412169169,
      "generation_ms": 7613.212800002657,
      "mean_inter_token_latency_ms": 155.3716897959726,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 30,
      "time_to_first_output_ms": 6420.845599997847,
      "ttft_ms": 6420.846100001654,
      "visible_generation_ms": 7613.21229999885
    },
    {
      "content_chunks": 24,
      "decode_tok_s": 44.75580857859987,
      "e2e_ms": 11736.917700000049,
      "effective_prefill_tok_s": 281.8514342098621,
      "generation_ms": 1206.5473000038764,
      "mean_inter_token_latency_ms": 22.343468518590303,
      "output_token_source": "usage",
      "output_tokens": 55,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 31,
      "time_to_first_output_ms": 10530.370399996173,
      "ttft_ms": 10530.371099994227,
      "visible_generation_ms": 1206.5466000058223
    }
  ],
  "requests": 32,
  "run_id": "benchmark-20260829T235745Z",
  "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": "2026-08-29T23:57:45Z",
  "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": 38326.56799999677
}
