{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 16,
  "concurrency": 8,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-k5",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T23:45:56Z",
  "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": 235,
    "content_chunks_s": null,
    "decode_tok_s_p50": 6.236136483958428,
    "decode_tok_s_p95": 44.44805673085673,
    "e2e_p50_ms": 7626.912499996251,
    "e2e_p95_ms": 19946.99160000164,
    "effective_prefill_tok_s_p50": 970.8039224660757,
    "effective_prefill_tok_s_p95": 2577.383821691005,
    "generation_p50_ms": 6292.594600003213,
    "generation_p95_ms": 12948.018000002776,
    "mean_inter_token_latency_ms_p50": 159.25786923076092,
    "mean_inter_token_latency_ms_p95": 390.6124878789838,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 671,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 16,
    "prompt_tokens": 47488,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 15,
    "throughput_tok_s": 28.648865533955135,
    "time_to_first_output_p50_ms": 3057.0532999990974,
    "time_to_first_output_p95_ms": 13735.934700001962,
    "ttft_p50_ms": 3057.0560000051046,
    "ttft_p95_ms": 13735.935099997732,
    "ttft_samples": 16
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 10,
      "decode_tok_s": 3.1468736422718466,
      "e2e_ms": 14814.9612000052,
      "effective_prefill_tok_s": 970.8039224660757,
      "generation_ms": 11757.701200003794,
      "mean_inter_token_latency_ms": 317.77570810821067,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 0,
      "time_to_first_output_ms": 3057.2600000014063,
      "ttft_ms": 4293.519099999685,
      "visible_generation_ms": 10521.442100005515
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 2.5600820020629023,
      "e2e_ms": 15947.265400005563,
      "effective_prefill_tok_s": 970.8695625296675,
      "generation_ms": 12890.212100006465,
      "mean_inter_token_latency_ms": 390.6124878789838,
      "output_token_source": "usage",
      "output_tokens": 34,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 3057.0532999990974,
      "ttft_ms": 3057.0560000051046,
      "visible_generation_ms": 12890.209400000458
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 3.423524795491422,
      "e2e_ms": 16024.94299999671,
      "effective_prefill_tok_s": 392.8970774337728,
      "generation_ms": 8470.801800001936,
      "mean_inter_token_latency_ms": 292.0966137931702,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 7554.141199994774,
      "ttft_ms": 7554.14199999359,
      "visible_generation_ms": 8470.80100000312
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 3.7843629812678277,
      "e2e_ms": 19887.07379999687,
      "effective_prefill_tok_s": 427.72389868986585,
      "generation_ms": 12948.018000002776,
      "mean_inter_token_latency_ms": 264.2452653061791,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 6939.055799994094,
      "ttft_ms": 6939.0563999986625,
      "visible_generation_ms": 12948.017399998207
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 2.9079534336841983,
      "e2e_ms": 15579.6782000034,
      "effective_prefill_tok_s": 603.3427914632065,
      "generation_ms": 10660.418300001766,
      "mean_inter_token_latency_ms": 343.88446129037953,
      "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": 4919.259900001634,
      "ttft_ms": 4919.260399998166,
      "visible_generation_ms": 10660.417800005234
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 6.279124572180628,
      "e2e_ms": 19946.99160000164,
      "effective_prefill_tok_s": 216.0755758397407,
      "generation_ms": 6211.056899999676,
      "mean_inter_token_latency_ms": 159.25786923076092,
      "output_token_source": "usage",
      "output_tokens": 40,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 13735.934700001962,
      "ttft_ms": 13735.935099997732,
      "visible_generation_ms": 6211.056500003906
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 2.9649419890260083,
      "e2e_ms": 13630.720200002543,
      "effective_prefill_tok_s": 2577.383821691005,
      "generation_ms": 12479.164900003525,
      "mean_inter_token_latency_ms": 337.2747270271223,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 6,
      "time_to_first_output_ms": 1151.5552999990177,
      "ttft_ms": 2377.7923000016017,
      "visible_generation_ms": 11252.927900000941
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 3.3550741229715246,
      "e2e_ms": 15875.578999999561,
      "effective_prefill_tok_s": 612.2735219741448,
      "generation_ms": 11028.07230000326,
      "mean_inter_token_latency_ms": 298.0560081081962,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 4847.506699996302,
      "ttft_ms": 6232.0516999971005,
      "visible_generation_ms": 9643.52730000246
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 6.236136483958428,
      "e2e_ms": 7626.912499996251,
      "effective_prefill_tok_s": 2447.462229884665,
      "generation_ms": 6414.227799999935,
      "mean_inter_token_latency_ms": 160.35569499999838,
      "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": 1212.6846999963163,
      "ttft_ms": 1212.6848999978392,
      "visible_generation_ms": 6414.227599998412
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 6.356678372380699,
      "e2e_ms": 7502.902699998231,
      "effective_prefill_tok_s": 2452.2681456169857,
      "generation_ms": 6292.594600003213,
      "mean_inter_token_latency_ms": 157.31486500008032,
      "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": 1210.308099995018,
      "ttft_ms": 1210.3084999980638,
      "visible_generation_ms": 6292.594200000167
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 8.953888694214113,
      "e2e_ms": 7265.316500001063,
      "effective_prefill_tok_s": 1262.3070149248501,
      "generation_ms": 4914.065999997547,
      "mean_inter_token_latency_ms": 111.68331818176243,
      "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": 2351.250500003516,
      "ttft_ms": 2351.250899999286,
      "visible_generation_ms": 4914.065600001777
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 10.982774890423373,
      "e2e_ms": 6606.016900004761,
      "effective_prefill_tok_s": 1141.6511337225945,
      "generation_ms": 4006.2735000028624,
      "mean_inter_token_latency_ms": 91.05167045461052,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 11,
      "time_to_first_output_ms": 2599.7434000018984,
      "ttft_ms": 2599.7443000014755,
      "visible_generation_ms": 4006.2726000032853
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 13.009018840541222,
      "e2e_ms": 7070.748199999798,
      "effective_prefill_tok_s": 919.6653707073818,
      "generation_ms": 3843.4874000013224,
      "mean_inter_token_latency_ms": 76.86974800002645,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 12,
      "time_to_first_output_ms": 3227.2607999984757,
      "ttft_ms": 3939.6619999970426,
      "visible_generation_ms": 3131.0862000027555
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 14.54608606406475,
      "e2e_ms": 6940.940100001171,
      "effective_prefill_tok_s": 785.4808002608914,
      "generation_ms": 3162.3627000008128,
      "mean_inter_token_latency_ms": 68.74701521740897,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 13,
      "time_to_first_output_ms": 3778.5774000003585,
      "ttft_ms": 3943.6381999985315,
      "visible_generation_ms": 2997.3019000026397
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 37.55907196940385,
      "e2e_ms": 3388.184600000386,
      "effective_prefill_tok_s": 1442.914243396443,
      "generation_ms": 1331.2363000004552,
      "mean_inter_token_latency_ms": 26.624726000009105,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 14,
      "time_to_first_output_ms": 2056.948299999931,
      "ttft_ms": 2056.9487000029767,
      "visible_generation_ms": 1331.2358999974094
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 44.44805673085673,
      "e2e_ms": 3450.6865000003017,
      "effective_prefill_tok_s": 1263.905892498266,
      "generation_ms": 1102.4104000025545,
      "mean_inter_token_latency_ms": 22.49817142862356,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 2348.276099997747,
      "ttft_ms": 2348.276500000793,
      "visible_generation_ms": 1102.4099999995087
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260829T234532Z",
  "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": "2026-08-29T23:45:32Z",
  "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": 23421.520800002327
}
