{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 16,
  "concurrency": 16,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-b12x-fixed-mtp5",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T19:47:17Z",
  "gpu": "dual-rtx-pro-6000-tp2",
  "manifest": null,
  "max_context_tokens": 524288,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 250,
    "content_chunks_s": null,
    "decode_tok_s_p50": 17.29133710275889,
    "decode_tok_s_p95": 45.896694831049075,
    "e2e_p50_ms": 24551.58239999946,
    "e2e_p95_ms": 32656.051600002684,
    "effective_prefill_tok_s_p50": 129.9448048058266,
    "effective_prefill_tok_s_p95": 1253.9836584968514,
    "generation_p50_ms": 2667.5384000009217,
    "generation_p95_ms": 14449.790899998334,
    "mean_inter_token_latency_ms_p50": 57.299187755104086,
    "mean_inter_token_latency_ms_p95": 321.10646444440744,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 780,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 16,
    "prompt_tokens": 47488,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 33,
    "throughput_tok_s": 23.851830885346413,
    "time_to_first_output_p50_ms": 21608.16650000197,
    "time_to_first_output_p95_ms": 31962.738199999876,
    "ttft_p50_ms": 21740.524099997856,
    "ttft_p95_ms": 31962.738599999284,
    "ttft_samples": 16
  },
  "model": "glm53-flash-tr3-4bpw-tp2-524k-fixed-mtp5",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 3.1685677301220556,
      "e2e_ms": 10697.265299997525,
      "effective_prefill_tok_s": 1191.1713175264956,
      "generation_ms": 8205.600199999026,
      "mean_inter_token_latency_ms": 315.6000076922702,
      "output_token_source": "usage",
      "output_tokens": 27,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 2491.6650999984995,
      "ttft_ms": 2491.6657999965537,
      "visible_generation_ms": 8205.599500000972
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 15.525328456302539,
      "e2e_ms": 27139.535700000124,
      "effective_prefill_tok_s": 121.14933591265529,
      "generation_ms": 2640.8459000012954,
      "mean_inter_token_latency_ms": 64.4108756097877,
      "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": 24498.68979999883,
      "ttft_ms": 24498.690199998236,
      "visible_generation_ms": 2640.8455000018876
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 17.452254371806905,
      "e2e_ms": 21865.431800000806,
      "effective_prefill_tok_s": 155.73699078227435,
      "generation_ms": 2807.6602000001003,
      "mean_inter_token_latency_ms": 57.299187755104086,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 19057.771600000706,
      "ttft_ms": 19057.772200001637,
      "visible_generation_ms": 2807.6595999991696
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 17.088552651998704,
      "e2e_ms": 23381.94889999795,
      "effective_prefill_tok_s": 143.0457469118598,
      "generation_ms": 2633.341799999471,
      "mean_inter_token_latency_ms": 58.51870666665491,
      "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": 20748.60709999848,
      "ttft_ms": 21740.524099997856,
      "visible_generation_ms": 1641.4248000000953
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 45.896694831049075,
      "e2e_ms": 32638.16809999844,
      "effective_prefill_tok_s": 92.85812690478475,
      "generation_ms": 675.4298999985622,
      "mean_inter_token_latency_ms": 21.7880612902762,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 31962.738199999876,
      "ttft_ms": 31962.738599999284,
      "visible_generation_ms": 675.4294999991544
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 26.280166784436272,
      "e2e_ms": 28471.333500001492,
      "effective_prefill_tok_s": 110.91588884906278,
      "generation_ms": 1712.3179000009259,
      "mean_inter_token_latency_ms": 38.05150888890947,
      "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": 26759.015600000566,
      "ttft_ms": 26886.306199998216,
      "visible_generation_ms": 1585.0273000032757
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 28.378745576298115,
      "e2e_ms": 32656.051600002684,
      "effective_prefill_tok_s": 96.73189657829745,
      "generation_ms": 1973.307800002658,
      "mean_inter_token_latency_ms": 35.23763928576175,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 6,
      "time_to_first_output_ms": 30682.743800000026,
      "ttft_ms": 30808.043500001077,
      "visible_generation_ms": 1848.008100001607
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 26.298632307496753,
      "e2e_ms": 24551.58239999946,
      "effective_prefill_tok_s": 129.9448048058266,
      "generation_ms": 1711.1155999991752,
      "mean_inter_token_latency_ms": 38.02479111109278,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 22840.466800000286,
      "ttft_ms": 22968.70249999847,
      "visible_generation_ms": 1582.8799000009894
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 18.99838802071182,
      "e2e_ms": 19618.145899999945,
      "effective_prefill_tok_s": 178.0387583898361,
      "generation_ms": 2947.618499998498,
      "mean_inter_token_latency_ms": 52.63604464283032,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 8,
      "time_to_first_output_ms": 16670.527400001447,
      "ttft_ms": 16798.526800001127,
      "visible_generation_ms": 2819.6190999988175
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 6.2010721087481535,
      "e2e_ms": 17532.76890000052,
      "effective_prefill_tok_s": 293.4344013159333,
      "generation_ms": 7418.0721000011545,
      "mean_inter_token_latency_ms": 161.26243695654685,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 9,
      "time_to_first_output_ms": 10114.696799999365,
      "ttft_ms": 10234.848799998872,
      "visible_generation_ms": 7297.920100001647
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 3.114231915979157,
      "e2e_ms": 16816.647899999225,
      "effective_prefill_tok_s": 1253.9836584968514,
      "generation_ms": 14449.790899998334,
      "mean_inter_token_latency_ms": 321.10646444440744,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 10,
      "time_to_first_output_ms": 2366.8570000008913,
      "ttft_ms": 10101.685699999507,
      "visible_generation_ms": 6714.962199999718
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 18.74387262803142,
      "e2e_ms": 32084.93879999878,
      "effective_prefill_tok_s": 100.8926675927563,
      "generation_ms": 2667.5384000009217,
      "mean_inter_token_latency_ms": 53.350768000018434,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 11,
      "time_to_first_output_ms": 29417.40039999786,
      "ttft_ms": 29545.36949999965,
      "visible_generation_ms": 2539.5692999991297
    },
    {
      "content_chunks": 23,
      "decode_tok_s": 16.979080050131596,
      "e2e_ms": 29545.176299998275,
      "effective_prefill_tok_s": 115.94137724270193,
      "generation_ms": 3946.0323999992397,
      "mean_inter_token_latency_ms": 58.89600597013791,
      "output_token_source": "usage",
      "output_tokens": 68,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 25599.143899999035,
      "ttft_ms": 25599.144399999204,
      "visible_generation_ms": 3946.0318999990704
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 17.29133710275889,
      "e2e_ms": 25598.604000002524,
      "effective_prefill_tok_s": 137.35547622699636,
      "generation_ms": 3990.437500000553,
      "mean_inter_token_latency_ms": 57.8324275362399,
      "output_token_source": "usage",
      "output_tokens": 70,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 13,
      "time_to_first_output_ms": 21608.16650000197,
      "ttft_ms": 21732.71250000107,
      "visible_generation_ms": 3865.891500001453
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 20.045451390591875,
      "e2e_ms": 20260.759699998744,
      "effective_prefill_tok_s": 166.12372809071087,
      "generation_ms": 2394.558199997846,
      "mean_inter_token_latency_ms": 49.88662916662179,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 14,
      "time_to_first_output_ms": 17866.201500000898,
      "ttft_ms": 18858.933600000455,
      "visible_generation_ms": 1401.8260999982886
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 16.85968066190771,
      "e2e_ms": 30718.56630000184,
      "effective_prefill_tok_s": 105.813026115428,
      "generation_ms": 2669.089700000768,
      "mean_inter_token_latency_ms": 59.313104444461516,
      "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": 28049.47660000107,
      "ttft_ms": 28177.728099999513,
      "visible_generation_ms": 2540.8382000023266
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260829T194645Z",
  "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": "2026-08-29T19:46:45Z",
  "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": 32701.892099998076
}
