{
  "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-vision",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T20:42:53Z",
  "gpu": "dual-rtx-pro-6000-tp2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 244,
    "content_chunks_s": null,
    "decode_tok_s_p50": 13.239749525358892,
    "decode_tok_s_p95": 64.10510439783606,
    "e2e_p50_ms": 21012.765300001774,
    "e2e_p95_ms": 28235.763299999235,
    "effective_prefill_tok_s_p50": 148.81147861953582,
    "effective_prefill_tok_s_p95": 2595.687101464856,
    "generation_p50_ms": 3539.55390000192,
    "generation_p95_ms": 13898.348300004727,
    "mean_inter_token_latency_ms_p50": 73.85440000007293,
    "mean_inter_token_latency_ms_p95": 248.184791071513,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 802,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 16,
    "prompt_tokens": 47488,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 45,
    "throughput_tok_s": 28.302338966696425,
    "time_to_first_output_p50_ms": 18277.911800003494,
    "time_to_first_output_p95_ms": 27496.937199997774,
    "ttft_p50_ms": 18454.304600003525,
    "ttft_p95_ms": 27631.59629999427,
    "ttft_samples": 16
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-fixed-mtp5-vision",
  "request_timings": [
    {
      "content_chunks": 15,
      "decode_tok_s": 13.540154682713725,
      "e2e_ms": 24655.993899999885,
      "effective_prefill_tok_s": 140.5903288434085,
      "generation_ms": 3545.0112000035006,
      "mean_inter_token_latency_ms": 73.85440000007293,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 0,
      "time_to_first_output_ms": 21110.982699996384,
      "ttft_ms": 22394.568199997593,
      "visible_generation_ms": 2261.4257000022917
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 13.239749525358892,
      "e2e_ms": 23494.614099996397,
      "effective_prefill_tok_s": 148.81147861953582,
      "generation_ms": 3549.9160999970627,
      "mean_inter_token_latency_ms": 75.53012978717155,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 1,
      "time_to_first_output_ms": 19944.697999999335,
      "ttft_ms": 20074.219299996912,
      "visible_generation_ms": 3420.3947999994853
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 21.983492220347888,
      "e2e_ms": 28177.341499998874,
      "effective_prefill_tok_s": 112.79893270623865,
      "generation_ms": 1865.0358000013512,
      "mean_inter_token_latency_ms": 45.48867804881344,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 26312.305699997523,
      "ttft_ms": 26312.307299995155,
      "visible_generation_ms": 1865.03420000372
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 19.371459004913973,
      "e2e_ms": 20073.30880000518,
      "effective_prefill_tok_s": 167.2084076352991,
      "generation_ms": 2323.0051999998977,
      "mean_inter_token_latency_ms": 51.622337777775506,
      "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": 17750.30360000528,
      "ttft_ms": 18454.304600003525,
      "visible_generation_ms": 1619.0042000016547
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 12.966137969416803,
      "e2e_ms": 20751.01999999606,
      "effective_prefill_tok_s": 182.33388714796104,
      "generation_ms": 4473.190099997737,
      "mean_inter_token_latency_ms": 77.1239672413403,
      "output_token_source": "usage",
      "output_tokens": 59,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 4,
      "time_to_first_output_ms": 16277.829899998324,
      "ttft_ms": 16411.615600001824,
      "visible_generation_ms": 4339.404399994237
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 4.029255764152993,
      "e2e_ms": 16393.72230000299,
      "effective_prefill_tok_s": 1189.4008673657993,
      "generation_ms": 13898.348300004727,
      "mean_inter_token_latency_ms": 248.184791071513,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 5,
      "time_to_first_output_ms": 2495.3739999982645,
      "ttft_ms": 3807.1807000014815,
      "visible_generation_ms": 12586.54160000151
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 7.4009128127232655,
      "e2e_ms": 16412.852000001294,
      "effective_prefill_tok_s": 335.5106178824721,
      "generation_ms": 7566.634200004046,
      "mean_inter_token_latency_ms": 135.1184678572151,
      "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": 8846.217799997248,
      "ttft_ms": 8963.432399999874,
      "visible_generation_ms": 7449.41960000142
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 17.026416606956314,
      "e2e_ms": 25586.588900005154,
      "effective_prefill_tok_s": 133.10860283815256,
      "generation_ms": 3289.006800005154,
      "mean_inter_token_latency_ms": 58.73226428580632,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 22297.5821,
      "ttft_ms": 23000.838800006022,
      "visible_generation_ms": 2585.7500999991316
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 5.8675691193872135,
      "e2e_ms": 9563.446099993598,
      "effective_prefill_tok_s": 961.3984278686883,
      "generation_ms": 6476.276499997766,
      "mean_inter_token_latency_ms": 170.42832894730964,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 8,
      "time_to_first_output_ms": 3087.169599995832,
      "ttft_ms": 8814.412599997013,
      "visible_generation_ms": 749.0334999965853
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 6.263201811650303,
      "e2e_ms": 18820.367899999837,
      "effective_prefill_tok_s": 273.87070641455864,
      "generation_ms": 7983.137299997907,
      "mean_inter_token_latency_ms": 159.66274599995813,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 9,
      "time_to_first_output_ms": 10837.23060000193,
      "ttft_ms": 16267.683600002783,
      "visible_generation_ms": 2552.684299997054
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 18.282514950081033,
      "e2e_ms": 21012.765300001774,
      "effective_prefill_tok_s": 162.3817880552106,
      "generation_ms": 2734.8534999982803,
      "mean_inter_token_latency_ms": 54.697069999965606,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 10,
      "time_to_first_output_ms": 18277.911800003494,
      "ttft_ms": 18416.12349999923,
      "visible_generation_ms": 2596.641800002544
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 6.817635347199243,
      "e2e_ms": 9504.10529999499,
      "effective_prefill_tok_s": 2595.687101464856,
      "generation_ms": 8360.670099995332,
      "mean_inter_token_latency_ms": 146.67842280693566,
      "output_token_source": "usage",
      "output_tokens": 58,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 11,
      "time_to_first_output_ms": 1143.435199999658,
      "ttft_ms": 2424.486699994304,
      "visible_generation_ms": 7079.618600000686
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 19.457461966904194,
      "e2e_ms": 25234.365500000422,
      "effective_prefill_tok_s": 129.48469397775307,
      "generation_ms": 2312.7374000032432,
      "mean_inter_token_latency_ms": 51.39416444451652,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 12,
      "time_to_first_output_ms": 22921.62809999718,
      "ttft_ms": 23055.675800002064,
      "visible_generation_ms": 2178.6896999983583
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 64.10510439783606,
      "e2e_ms": 28183.310000000347,
      "effective_prefill_tok_s": 107.93929441713385,
      "generation_ms": 686.3728000025731,
      "mean_inter_token_latency_ms": 15.599381818240296,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 13,
      "time_to_first_output_ms": 27496.937199997774,
      "ttft_ms": 27631.59629999427,
      "visible_generation_ms": 551.7137000060757
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 12.713466519036649,
      "e2e_ms": 27969.490299998142,
      "effective_prefill_tok_s": 121.49028762925715,
      "generation_ms": 3539.55390000192,
      "mean_inter_token_latency_ms": 78.656753333376,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 14,
      "time_to_first_output_ms": 24429.936399996222,
      "ttft_ms": 24556.772399999318,
      "visible_generation_ms": 3412.7178999988246
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 36.60826208950212,
      "e2e_ms": 28235.763299999235,
      "effective_prefill_tok_s": 110.45795936942406,
      "generation_ms": 1365.811899995606,
      "mean_inter_token_latency_ms": 27.31623799991212,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 15,
      "time_to_first_output_ms": 26869.95140000363,
      "ttft_ms": 27575.08919999964,
      "visible_generation_ms": 660.6740999995964
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260829T204225Z",
  "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": "2026-08-29T20:42:25Z",
  "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": 28336.88059999986
}
