{
  "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-nospec-xgrammar-fixed",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-31T08:14:15Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 524288,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 184,
    "content_chunks_s": null,
    "decode_tok_s_p50": 43.631721675033454,
    "decode_tok_s_p95": 47.57521537932,
    "e2e_p50_ms": 60467.522899998585,
    "e2e_p95_ms": 63892.73940000567,
    "effective_prefill_tok_s_p50": 3332.0032428002287,
    "effective_prefill_tok_s_p95": 3355.064206453433,
    "generation_p50_ms": 962.6023999880999,
    "generation_p95_ms": 1054.4879999943078,
    "mean_inter_token_latency_ms_p50": 22.919104761621426,
    "mean_inter_token_latency_ms_p95": 23.43306666654017,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 195,
    "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": 0.6391785140854613,
    "time_to_first_output_p50_ms": 59466.62880000076,
    "time_to_first_output_p95_ms": 63409.29439998581,
    "ttft_p50_ms": 59466.62910000305,
    "ttft_p95_ms": 63409.2951999919,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-nospec-fp8-tp2-524k-vision-xgfix",
  "request_timings": [
    {
      "content_chunks": 21,
      "decode_tok_s": 47.57521537932,
      "e2e_ms": 63892.73940000567,
      "effective_prefill_tok_s": 3124.76273194493,
      "generation_ms": 483.4450000198558,
      "mean_inter_token_latency_ms": 21.019347826950252,
      "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": 63409.29439998581,
      "ttft_ms": 63409.2951999919,
      "visible_generation_ms": 483.4442000137642
    },
    {
      "content_chunks": 41,
      "decode_tok_s": 43.58161607047539,
      "e2e_ms": 60020.70110000204,
      "effective_prefill_tok_s": 3355.064206453433,
      "generation_ms": 963.7091000040527,
      "mean_inter_token_latency_ms": 22.945454762001255,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198140,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 59056.99199999799,
      "ttft_ms": 59056.9929999765,
      "visible_generation_ms": 963.7081000255421
    },
    {
      "content_chunks": 41,
      "decode_tok_s": 43.631721675033454,
      "e2e_ms": 60159.10940000322,
      "effective_prefill_tok_s": 3347.190738804055,
      "generation_ms": 962.6023999880999,
      "mean_inter_token_latency_ms": 22.919104761621426,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198142,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 59196.50700001512,
      "ttft_ms": 59196.50740001816,
      "visible_generation_ms": 962.6019999850541
    },
    {
      "content_chunks": 37,
      "decode_tok_s": 43.78609616340028,
      "e2e_ms": 60467.522899998585,
      "effective_prefill_tok_s": 3324.5823057650437,
      "generation_ms": 867.8553999925498,
      "mean_inter_token_latency_ms": 22.83829999980394,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198144,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 59599.667500006035,
      "ttft_ms": 59599.66790000908,
      "visible_generation_ms": 867.854999989504
    },
    {
      "content_chunks": 44,
      "decode_tok_s": 42.67473883082872,
      "e2e_ms": 60521.11679999507,
      "effective_prefill_tok_s": 3332.0032428002287,
      "generation_ms": 1054.4879999943078,
      "mean_inter_token_latency_ms": 23.43306666654017,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198143,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 59466.62880000076,
      "ttft_ms": 59466.62910000305,
      "visible_generation_ms": 1054.4876999920234
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260831T080910Z",
  "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": "2026-08-31T08:09:10Z",
  "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": 305079.09089999157
}