{
  "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-k3",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T23:57:37Z",
  "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": 78,
    "content_chunks_s": null,
    "decode_tok_s_p50": 66.77897249590553,
    "decode_tok_s_p95": 75.81006848663463,
    "e2e_p50_ms": 1577.3263000010047,
    "e2e_p95_ms": 9524.021899997024,
    "effective_prefill_tok_s_p50": 3068.4899782934494,
    "effective_prefill_tok_s_p95": 3167.548930788445,
    "generation_p50_ms": 569.0414000055171,
    "generation_p95_ms": 966.0603999946034,
    "mean_inter_token_latency_ms_p50": 14.974773684355712,
    "mean_inter_token_latency_ms_p95": 25.42264210512114,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 208,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 14840,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 9,
    "throughput_tok_s": 13.185468196610628,
    "time_to_first_output_p50_ms": 967.2510000018519,
    "time_to_first_output_p95_ms": 8557.96150000242,
    "ttft_p50_ms": 1008.2886000018334,
    "ttft_p95_ms": 9097.287299999152,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-dflash2-k3-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 13,
      "decode_tok_s": 39.33501466389915,
      "e2e_ms": 9524.021899997024,
      "effective_prefill_tok_s": 346.81156254315476,
      "generation_ms": 966.0603999946034,
      "mean_inter_token_latency_ms": 25.42264210512114,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 0,
      "time_to_first_output_ms": 8557.96150000242,
      "ttft_ms": 9097.287299999152,
      "visible_generation_ms": 426.73459999787156
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 66.77897249590553,
      "e2e_ms": 1577.3263000010047,
      "effective_prefill_tok_s": 2943.612465101166,
      "generation_ms": 569.0414000055171,
      "mean_inter_token_latency_ms": 14.974773684355712,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 1008.2848999954876,
      "ttft_ms": 1008.2886000018334,
      "visible_generation_ms": 569.0376999991713
    },
    {
      "content_chunks": 23,
      "decode_tok_s": 70.06711115482464,
      "e2e_ms": 1875.4808000012417,
      "effective_prefill_tok_s": 3085.0120386143335,
      "generation_ms": 913.4100000010221,
      "mean_inter_token_latency_ms": 14.27203125001597,
      "output_token_source": "usage",
      "output_tokens": 65,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 962.0708000002196,
      "ttft_ms": 1065.961000000243,
      "visible_generation_ms": 809.5198000009987
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 75.81006848663463,
      "e2e_ms": 1362.976800002798,
      "effective_prefill_tok_s": 3068.4899782934494,
      "generation_ms": 395.72580000094604,
      "mean_inter_token_latency_ms": 13.190860000031535,
      "output_token_source": "usage",
      "output_tokens": 31,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 967.2510000018519,
      "ttft_ms": 967.2518000006676,
      "visible_generation_ms": 395.7250000021304
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 66.52446558205466,
      "e2e_ms": 1433.0601999972714,
      "effective_prefill_tok_s": 3167.548930788445,
      "generation_ms": 496.05809999775374,
      "mean_inter_token_latency_ms": 15.032063636295568,
      "output_token_source": "usage",
      "output_tokens": 34,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 4,
      "time_to_first_output_ms": 937.0020999995177,
      "ttft_ms": 1003.4938000026159,
      "visible_generation_ms": 429.5663999946555
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260829T235722Z",
  "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": "2026-08-29T23:57:22Z",
  "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": 15774.942299998656
}