{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 2,
  "concurrency": 2,
  "context_distribution": {
    "240000": 2
  },
  "context_seed": 0,
  "context_tokens": 240000,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-k5",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-30T00:11:59Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 1048576,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 31,
    "content_chunks_s": null,
    "decode_tok_s_p50": 4.248900262649669,
    "decode_tok_s_p95": 68.87983941326854,
    "e2e_p50_ms": 43763.30450000387,
    "e2e_p95_ms": 71369.2979999978,
    "effective_prefill_tok_s_p50": 2803.063354113972,
    "effective_prefill_tok_s_p95": 5296.599437152732,
    "generation_p50_ms": 682.3476999998093,
    "generation_p95_ms": 6354.5855000047595,
    "mean_inter_token_latency_ms_p50": 14.518036170208708,
    "mean_inter_token_latency_ms_p95": 235.3550185186948,
    "output_token_sources": {
      "usage": 2
    },
    "output_tokens": 76,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 2,
    "prompt_tokens": 396279,
    "prompt_tokens_p50": 198139.0,
    "prompt_tokens_p95": 198140.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 1.0647729772667756,
    "time_to_first_output_p50_ms": 37408.71899999911,
    "time_to_first_output_p95_ms": 70686.95029999799,
    "ttft_p50_ms": 37408.720500003255,
    "ttft_p95_ms": 70686.95119999757,
    "ttft_samples": 2
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 4.248900262649669,
      "e2e_ms": 43763.30450000387,
      "effective_prefill_tok_s": 5296.599437152732,
      "generation_ms": 6354.5855000047595,
      "mean_inter_token_latency_ms": 235.3550185186948,
      "output_token_source": "usage",
      "output_tokens": 28,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198139,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 37408.71899999911,
      "ttft_ms": 37408.720500003255,
      "visible_generation_ms": 6354.584000000614
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 68.87983941326854,
      "e2e_ms": 71369.2979999978,
      "effective_prefill_tok_s": 2803.063354113972,
      "generation_ms": 682.3476999998093,
      "mean_inter_token_latency_ms": 14.518036170208708,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198140,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 70686.95029999799,
      "ttft_ms": 70686.95119999757,
      "visible_generation_ms": 682.3468000002322
    }
  ],
  "requests": 2,
  "run_id": "benchmark-20260830T001048Z",
  "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": "2026-08-30T00:10:48Z",
  "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": 71376.71749999572
}