{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 5,
  "concurrency": 1,
  "context_distribution": {
    "4096": 5
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-k5-bt4096",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-30T00:22:25Z",
  "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": 71,
    "content_chunks_s": null,
    "decode_tok_s_p50": 62.495185917344735,
    "decode_tok_s_p95": 84.4673676495519,
    "e2e_p50_ms": 1618.787200000952,
    "e2e_p95_ms": 4113.881000004767,
    "effective_prefill_tok_s_p50": 3034.4992299250016,
    "effective_prefill_tok_s_p95": 3174.2161247547683,
    "generation_p50_ms": 640.0493000037386,
    "generation_p95_ms": 832.2955000039656,
    "mean_inter_token_latency_ms_p50": 16.001232500093465,
    "mean_inter_token_latency_ms_p95": 23.1193194445546,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 198,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 14840,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 6,
    "throughput_tok_s": 19.37133767373352,
    "time_to_first_output_p50_ms": 978.0855999997584,
    "time_to_first_output_p95_ms": 3281.5855000008014,
    "ttft_p50_ms": 978.7389999983134,
    "ttft_p95_ms": 3739.3919000023743,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision-bt4096",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 43.253868367459,
      "e2e_ms": 4113.881000004767,
      "effective_prefill_tok_s": 904.4408564089753,
      "generation_ms": 832.2955000039656,
      "mean_inter_token_latency_ms": 23.1193194445546,
      "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": 3281.5855000008014,
      "ttft_ms": 3739.3919000023743,
      "visible_generation_ms": 374.48910000239266
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 69.00209898533126,
      "e2e_ms": 1659.2242999977316,
      "effective_prefill_tok_s": 3034.4992299250016,
      "generation_ms": 681.1386999979732,
      "mean_inter_token_latency_ms": 14.492312765914322,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 978.0855999997584,
      "ttft_ms": 978.0881999977282,
      "visible_generation_ms": 681.1361000000034
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 62.495185917344735,
      "e2e_ms": 1618.787200000952,
      "effective_prefill_tok_s": 3032.4768255203467,
      "generation_ms": 640.0493000037386,
      "mean_inter_token_latency_ms": 16.001232500093465,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 978.7378999972134,
      "ttft_ms": 978.7389999983134,
      "visible_generation_ms": 640.0482000026386
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 84.4673676495519,
      "e2e_ms": 1491.4619000046514,
      "effective_prefill_tok_s": 3174.2161247547683,
      "generation_ms": 556.4279000027454,
      "mean_inter_token_latency_ms": 11.838891489420115,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 3,
      "time_to_first_output_ms": 935.034000001906,
      "ttft_ms": 1014.2282000015257,
      "visible_generation_ms": 477.2337000031257
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 62.167303566998335,
      "e2e_ms": 1336.1850999935996,
      "effective_prefill_tok_s": 3071.777864926875,
      "generation_ms": 369.96939999517053,
      "mean_inter_token_latency_ms": 16.085626086746544,
      "output_token_source": "usage",
      "output_tokens": 24,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 966.215699998429,
      "ttft_ms": 966.2164999972447,
      "visible_generation_ms": 369.9685999963549
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260830T002215Z",
  "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": "2026-08-30T00:22:15Z",
  "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": 10221.28689999954
}