{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 2,
  "concurrency": 2,
  "context_distribution": {
    "250000": 2
  },
  "context_seed": 0,
  "context_tokens": 250000,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-xgrammar-fixed",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-31T08:58:30Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 524288,
  "max_tokens": 8192,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 37,
    "content_chunks_s": null,
    "decode_tok_s_p50": 4.76456476115068,
    "decode_tok_s_p95": 65.99331487719655,
    "e2e_p50_ms": 83820.40090000373,
    "e2e_p95_ms": 133940.17970000277,
    "effective_prefill_tok_s_p50": 1550.4213853313772,
    "effective_prefill_tok_s_p95": 2754.871070820435,
    "generation_p50_ms": 666.7342000000644,
    "generation_p95_ms": 8815.075899998192,
    "mean_inter_token_latency_ms_p50": 15.153050000001464,
    "mean_inter_token_latency_ms_p95": 209.88275952376648,
    "output_token_sources": {
      "usage": 2
    },
    "output_tokens": 88,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 2,
    "prompt_tokens": 413260,
    "prompt_tokens_p50": 206630.0,
    "prompt_tokens_p95": 206630.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 0.6569727426899759,
    "time_to_first_output_p50_ms": 75005.32500000554,
    "time_to_first_output_p95_ms": 133273.4455000027,
    "ttft_p50_ms": 75005.32680001925,
    "ttft_p95_ms": 133273.44640000956,
    "ttft_samples": 2
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-524k-vision-xgfix",
  "request_timings": [
    {
      "content_chunks": 17,
      "decode_tok_s": 4.76456476115068,
      "e2e_ms": 83820.40090000373,
      "effective_prefill_tok_s": 2754.871070820435,
      "generation_ms": 8815.075899998192,
      "mean_inter_token_latency_ms": 209.88275952376648,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 250000,
      "prompt_tokens": 206630,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 75005.32500000554,
      "ttft_ms": 75005.32680001925,
      "visible_generation_ms": 8815.074099984486
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 65.99331487719655,
      "e2e_ms": 133940.17970000277,
      "effective_prefill_tok_s": 1550.4213853313772,
      "generation_ms": 666.7342000000644,
      "mean_inter_token_latency_ms": 15.153050000001464,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 250000,
      "prompt_tokens": 206630,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 133273.4455000027,
      "ttft_ms": 133273.44640000956,
      "visible_generation_ms": 666.7332999932114
    }
  ],
  "requests": 2,
  "run_id": "benchmark-20260831T085616Z",
  "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": "2026-08-31T08:56:16Z",
  "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": 133947.71849998506
}