{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 5,
  "concurrency": 1,
  "context_distribution": {
    "65536": 5
  },
  "context_seed": 0,
  "context_tokens": 65536,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-k5",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T23:28:19Z",
  "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": 94,
    "content_chunks_s": null,
    "decode_tok_s_p50": 67.2015437962451,
    "decode_tok_s_p95": 82.46212669068544,
    "e2e_p50_ms": 11726.896100000886,
    "e2e_p95_ms": 12028.827100002673,
    "effective_prefill_tok_s_p50": 4554.523728820982,
    "effective_prefill_tok_s_p95": 4585.746825006739,
    "generation_p50_ms": 582.0854000048712,
    "generation_p95_ms": 889.7454999969341,
    "mean_inter_token_latency_ms_p50": 14.880610526329532,
    "mean_inter_token_latency_ms_p95": 16.787650943338377,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 240,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 252752,
    "prompt_tokens_p50": 50551.0,
    "prompt_tokens_p95": 50553.0,
    "reasoning_chunks": 6,
    "throughput_tok_s": 4.069330961754462,
    "time_to_first_output_p50_ms": 11099.074899997504,
    "time_to_first_output_p95_ms": 11161.432900000364,
    "ttft_p50_ms": 11158.474099996965,
    "ttft_p95_ms": 11233.877500002563,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 15,
      "decode_tok_s": 82.46212669068544,
      "e2e_ms": 11660.5825000006,
      "effective_prefill_tok_s": 4562.622487848057,
      "generation_ms": 582.0854000048712,
      "mean_inter_token_latency_ms": 12.12677916676815,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 65536,
      "prompt_tokens": 50547,
      "reasoning_chunks": 3,
      "request_index": 0,
      "time_to_first_output_ms": 11078.49709999573,
      "ttft_ms": 11158.474099996965,
      "visible_generation_ms": 502.1084000036353
    },
    {
      "content_chunks": 26,
      "decode_tok_s": 59.576045618854934,
      "e2e_ms": 11912.687500000175,
      "effective_prefill_tok_s": 4585.746825006739,
      "generation_ms": 889.6192999964114,
      "mean_inter_token_latency_ms": 16.785269811253045,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 65536,
      "prompt_tokens": 50549,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 11023.068200003763,
      "ttft_ms": 11023.07130000554,
      "visible_generation_ms": 889.6161999946344
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 79.07883092764845,
      "e2e_ms": 11642.83609999984,
      "effective_prefill_tok_s": 4554.523728820982,
      "generation_ms": 543.7612000023364,
      "mean_inter_token_latency_ms": 12.645609302379917,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 65536,
      "prompt_tokens": 50551,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 11099.074899997504,
      "ttft_ms": 11099.075799997081,
      "visible_generation_ms": 543.7603000027593
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 67.2015437962451,
      "e2e_ms": 11726.896100000886,
      "effective_prefill_tok_s": 4529.167576682592,
      "generation_ms": 565.4632000005222,
      "mean_inter_token_latency_ms": 14.880610526329532,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 65536,
      "prompt_tokens": 50552,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 11161.432900000364,
      "ttft_ms": 11161.43340000417,
      "visible_generation_ms": 565.462699996715
    },
    {
      "content_chunks": 22,
      "decode_tok_s": 59.56759545306228,
      "e2e_ms": 12028.827100002673,
      "effective_prefill_tok_s": 4538.345423376192,
      "generation_ms": 889.7454999969341,
      "mean_inter_token_latency_ms": 16.787650943338377,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 65536,
      "prompt_tokens": 50553,
      "reasoning_chunks": 3,
      "request_index": 4,
      "time_to_first_output_ms": 11139.081600005738,
      "ttft_ms": 11233.877500002563,
      "visible_generation_ms": 794.9496000001091
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260829T232720Z",
  "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": "2026-08-29T23:27:20Z",
  "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": 58977.75389999879
}