{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 5,
  "concurrency": 1,
  "context_distribution": {
    "240000": 5
  },
  "context_seed": 0,
  "context_tokens": 240000,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-xgrammar-fixed",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-31T08:55:37Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 524288,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 89,
    "content_chunks_s": null,
    "decode_tok_s_p50": 70.49651997462487,
    "decode_tok_s_p95": 77.23243496660108,
    "e2e_p50_ms": 60229.799799999455,
    "e2e_p95_ms": 60553.812400001334,
    "effective_prefill_tok_s_p50": 3320.46102649467,
    "effective_prefill_tok_s_p95": 3358.0880333884515,
    "generation_p50_ms": 579.2350000119768,
    "generation_p95_ms": 662.0830000028946,
    "mean_inter_token_latency_ms_p50": 14.18509736877719,
    "mean_inter_token_latency_ms_p95": 18.288620000073156,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 210,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 990708,
    "prompt_tokens_p50": 198142.0,
    "prompt_tokens_p95": 198144.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 0.697535193824288,
    "time_to_first_output_p50_ms": 59673.038900014944,
    "time_to_first_output_p95_ms": 59974.57739998936,
    "ttft_p50_ms": 59673.039299988886,
    "ttft_p95_ms": 59974.5777999924,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-524k-vision-xgfix",
  "request_timings": [
    {
      "content_chunks": 16,
      "decode_tok_s": 70.49651997462487,
      "e2e_ms": 59542.54940000828,
      "effective_prefill_tok_s": 3358.0880333884515,
      "generation_ms": 539.0337000135332,
      "mean_inter_token_latency_ms": 14.18509736877719,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198139,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 59003.51569999475,
      "ttft_ms": 59003.5166000016,
      "visible_generation_ms": 539.0328000066802
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 54.678811194939804,
      "e2e_ms": 60205.44050002354,
      "effective_prefill_tok_s": 3326.4311761109334,
      "generation_ms": 640.1017000025604,
      "mean_inter_token_latency_ms": 18.288620000073156,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198140,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 59565.33880002098,
      "ttft_ms": 59565.339700027835,
      "visible_generation_ms": 640.1007999957073
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 77.23243496660108,
      "e2e_ms": 60229.799799999455,
      "effective_prefill_tok_s": 3320.46102649467,
      "generation_ms": 556.7608999845106,
      "mean_inter_token_latency_ms": 12.947927906616526,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198142,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 59673.038900014944,
      "ttft_ms": 59673.039299988886,
      "visible_generation_ms": 556.7605000105686
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 74.23584555337797,
      "e2e_ms": 60553.812400001334,
      "effective_prefill_tok_s": 3303.799853036316,
      "generation_ms": 579.2350000119768,
      "mean_inter_token_latency_ms": 13.470581395627368,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198144,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 59974.57739998936,
      "ttft_ms": 59974.5777999924,
      "visible_generation_ms": 579.234600008931
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 69.47769388399777,
      "e2e_ms": 60512.009600002784,
      "effective_prefill_tok_s": 3310.6640434877386,
      "generation_ms": 662.0830000028946,
      "mean_inter_token_latency_ms": 14.3931086957151,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198143,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 59849.92659999989,
      "ttft_ms": 59849.926900002174,
      "visible_generation_ms": 662.0827000006102
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260831T085036Z",
  "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": "2026-08-31T08:50:36Z",
  "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": 301060.07820001105
}