{
  "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-b12x-fixed-mtp5-vision",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T22:29:40Z",
  "gpu": "dual-rtx-pro-6000-max-q-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": 20.880835528182676,
    "decode_tok_s_p95": 55.21585088147601,
    "e2e_p50_ms": 9263.900799996918,
    "e2e_p95_ms": 10372.93349999527,
    "effective_prefill_tok_s_p50": 434.1484812378148,
    "effective_prefill_tok_s_p95": 1747.7928962006367,
    "generation_p50_ms": 2343.9376999958768,
    "generation_p95_ms": 4339.581699998234,
    "mean_inter_token_latency_ms_p50": 45.959562745017195,
    "mean_inter_token_latency_ms_p95": 79.70654888874076,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 735,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 16,
    "prompt_tokens": 47488,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 30,
    "throughput_tok_s": 38.817030965474935,
    "time_to_first_output_p50_ms": 6703.957799996715,
    "time_to_first_output_p95_ms": 8744.29180000152,
    "ttft_p50_ms": 7627.521800000977,
    "ttft_p95_ms": 8880.018299998483,
    "ttft_samples": 16
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-fixed-mtp5-vision",
  "request_timings": [
    {
      "content_chunks": 16,
      "decode_tok_s": 18.2551093109987,
      "e2e_ms": 4827.708199998597,
      "effective_prefill_tok_s": 1256.2192481822456,
      "generation_ms": 2465.0632999982918,
      "mean_inter_token_latency_ms": 54.77918444440648,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 0,
      "time_to_first_output_ms": 2362.644900000305,
      "ttft_ms": 3701.0450000016135,
      "visible_generation_ms": 1126.6631999969832
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 55.21585088147601,
      "e2e_ms": 4827.12029999675,
      "effective_prefill_tok_s": 726.6353144429986,
      "generation_ms": 742.5404000023264,
      "mean_inter_token_latency_ms": 18.110741463471374,
      "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": 4084.5798999944236,
      "ttft_ms": 4084.5807999940007,
      "visible_generation_ms": 742.5395000027493
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 17.203862252133828,
      "e2e_ms": 9510.189000000537,
      "effective_prefill_tok_s": 434.1484812378148,
      "generation_ms": 2673.8181999971857,
      "mean_inter_token_latency_ms": 58.126482608634475,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 6836.370800003351,
      "ttft_ms": 8155.420900002355,
      "visible_generation_ms": 1354.7680999981822
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 15.399628692949014,
      "e2e_ms": 4880.036599999585,
      "effective_prefill_tok_s": 1747.7928962006367,
      "generation_ms": 3181.8949000007706,
      "mean_inter_token_latency_ms": 64.93663061226063,
      "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": 1698.1416999988141,
      "ttft_ms": 1698.1436999994912,
      "visible_generation_ms": 3181.8929000000935
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 13.888409752543943,
      "e2e_ms": 10195.642899998347,
      "effective_prefill_tok_s": 481.54428005689425,
      "generation_ms": 4032.1390999961295,
      "mean_inter_token_latency_ms": 72.00248392850231,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 4,
      "time_to_first_output_ms": 6163.503800002218,
      "ttft_ms": 7491.753399997833,
      "visible_generation_ms": 2703.889500000514
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 30.318398306669113,
      "e2e_ms": 10195.555799997237,
      "effective_prefill_tok_s": 339.42142690154554,
      "generation_ms": 1451.2639999957173,
      "mean_inter_token_latency_ms": 32.9832727271754,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 5,
      "time_to_first_output_ms": 8744.29180000152,
      "ttft_ms": 8880.018299998483,
      "visible_generation_ms": 1315.5374999987544
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 48.641781612239186,
      "e2e_ms": 9470.170000000508,
      "effective_prefill_tok_s": 343.20741926556764,
      "generation_ms": 822.3383000004105,
      "mean_inter_token_latency_ms": 20.558457500010263,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 8647.831700000097,
      "ttft_ms": 8647.832000002381,
      "visible_generation_ms": 822.3379999981262
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 23.406852958127715,
      "e2e_ms": 4729.673499998171,
      "effective_prefill_tok_s": 1011.1308885766585,
      "generation_ms": 1794.3462999974145,
      "mean_inter_token_latency_ms": 42.72253095231939,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 2935.3272000007564,
      "ttft_ms": 3921.8076999968616,
      "visible_generation_ms": 807.8658000013093
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 12.904469571346654,
      "e2e_ms": 10372.93349999527,
      "effective_prefill_tok_s": 491.9321959646806,
      "generation_ms": 4339.581699998234,
      "mean_inter_token_latency_ms": 77.49253035711133,
      "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": 6033.351799997035,
      "ttft_ms": 7364.681699997163,
      "visible_generation_ms": 3008.2517999981064
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 14.844080512728988,
      "e2e_ms": 9263.900799996918,
      "effective_prefill_tok_s": 442.723550557173,
      "generation_ms": 2559.943000000203,
      "mean_inter_token_latency_ms": 67.36692105263693,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 9,
      "time_to_first_output_ms": 6703.957799996715,
      "ttft_ms": 8028.554800002894,
      "visible_generation_ms": 1235.3459999940242
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 21.758257482735022,
      "e2e_ms": 10371.505600000091,
      "effective_prefill_tok_s": 369.7259290698048,
      "generation_ms": 2343.9376999958768,
      "mean_inter_token_latency_ms": 45.959562745017195,
      "output_token_source": "usage",
      "output_tokens": 52,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 10,
      "time_to_first_output_ms": 8027.567900004215,
      "ttft_ms": 8698.586799997429,
      "visible_generation_ms": 1672.9188000026625
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 24.875858582483904,
      "e2e_ms": 10254.30729999789,
      "effective_prefill_tok_s": 343.2673215316711,
      "generation_ms": 1607.9847000000882,
      "mean_inter_token_latency_ms": 40.199617500002205,
      "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": 8646.322599997802,
      "ttft_ms": 8646.322900000087,
      "visible_generation_ms": 1607.9843999978038
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 12.546020545888403,
      "e2e_ms": 9210.786099996767,
      "effective_prefill_tok_s": 527.7390715779168,
      "generation_ms": 3586.794699993334,
      "mean_inter_token_latency_ms": 79.70654888874076,
      "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": 5623.991400003433,
      "ttft_ms": 6315.785699996923,
      "visible_generation_ms": 2895.0003999998444
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 20.880835528182676,
      "e2e_ms": 9365.352300002996,
      "effective_prefill_tok_s": 428.720748277184,
      "generation_ms": 2442.4310000031255,
      "mean_inter_token_latency_ms": 47.89080392162991,
      "output_token_source": "usage",
      "output_tokens": 52,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 13,
      "time_to_first_output_ms": 6922.921299999871,
      "ttft_ms": 8256.252300001506,
      "visible_generation_ms": 1109.1000000014901
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 50.40649876158213,
      "e2e_ms": 8685.802899999544,
      "effective_prefill_tok_s": 370.4773167087393,
      "generation_ms": 674.516200000653,
      "mean_inter_token_latency_ms": 19.83871176472509,
      "output_token_source": "usage",
      "output_tokens": 35,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 14,
      "time_to_first_output_ms": 8011.286699998891,
      "ttft_ms": 8011.287199995422,
      "visible_generation_ms": 674.5157000041218
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 36.96525032072528,
      "e2e_ms": 8736.67139999452,
      "effective_prefill_tok_s": 389.11721993684193,
      "generation_ms": 1109.1497999950661,
      "mean_inter_token_latency_ms": 27.052434146221124,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 7627.521599999454,
      "ttft_ms": 7627.521800000977,
      "visible_generation_ms": 1109.1495999935432
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260829T222921Z",
  "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": "2026-08-29T22:29:21Z",
  "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": 18934.98759999784
}
