{
  "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": "2026-08-31T08:49:53Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 524288,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 94,
    "content_chunks_s": null,
    "decode_tok_s_p50": 64.51235113835298,
    "decode_tok_s_p95": 74.30733028894842,
    "e2e_p50_ms": 60456.50639999076,
    "e2e_p95_ms": 64455.95979999052,
    "effective_prefill_tok_s_p50": 3313.3664472996734,
    "effective_prefill_tok_s_p95": 3323.431556971435,
    "generation_p50_ms": 584.310700010974,
    "generation_p95_ms": 996.8988999899011,
    "mean_inter_token_latency_ms_p50": 15.50090769216275,
    "mean_inter_token_latency_ms_p95": 16.932252173200656,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 213,
    "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": 0.6955765758281852,
    "time_to_first_output_p50_ms": 59801.11259999103,
    "time_to_first_output_p95_ms": 64066.51800000691,
    "ttft_p50_ms": 59801.112999994075,
    "ttft_p95_ms": 64066.51889998466,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-524k-vision-xgfix",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 59.05888890449786,
      "e2e_ms": 64455.95979999052,
      "effective_prefill_tok_s": 3092.707488800603,
      "generation_ms": 389.44179998361506,
      "mean_inter_token_latency_ms": 16.932252173200656,
      "output_token_source": "usage",
      "output_tokens": 24,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198139,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 64066.51800000691,
      "ttft_ms": 64066.51889998466,
      "visible_generation_ms": 389.44090000586584
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 64.51235113835298,
      "e2e_ms": 60223.63590000896,
      "effective_prefill_tok_s": 3323.431556971435,
      "generation_ms": 604.5353999943472,
      "mean_inter_token_latency_ms": 15.50090769216275,
      "output_token_source": "usage",
      "output_tokens": 40,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198140,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 59619.10050001461,
      "ttft_ms": 59619.10140002146,
      "visible_generation_ms": 604.5344999874942
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 73.59098506871844,
      "e2e_ms": 60266.07280000462,
      "effective_prefill_tok_s": 3319.9757015890973,
      "generation_ms": 584.310700010974,
      "mean_inter_token_latency_ms": 13.588620930487766,
      "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": 59681.76209999365,
      "ttft_ms": 59681.762499996694,
      "visible_generation_ms": 584.3103000079282
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 74.30733028894842,
      "e2e_ms": 60456.50639999076,
      "effective_prefill_tok_s": 3306.9083356917868,
      "generation_ms": 538.3048999938183,
      "mean_inter_token_latency_ms": 13.457622499845456,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198144,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 59918.20149999694,
      "ttft_ms": 59918.201899999985,
      "visible_generation_ms": 538.3044999907725
    },
    {
      "content_chunks": 30,
      "decode_tok_s": 63.19597704505262,
      "e2e_ms": 60798.01149998093,
      "effective_prefill_tok_s": 3313.3664472996734,
      "generation_ms": 996.8988999899011,
      "mean_inter_token_latency_ms": 15.823792063331764,
      "output_token_source": "usage",
      "output_tokens": 64,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198143,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 59801.11259999103,
      "ttft_ms": 59801.112999994075,
      "visible_generation_ms": 996.8984999868553
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260831T084446Z",
  "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": "2026-08-31T08:44:46Z",
  "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": 306220.7776999858
}