{
  "base_url": "http://127.0.0.1:39267/v1",
  "completed": 8,
  "concurrency": 4,
  "context_distribution": {
    "4096": 8
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang-5f55db35-dflash2-k4",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:09:26Z",
  "gpu": "RTX PRO 6000 Blackwell Max-Q TP1",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 512,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 136,
    "content_chunks_s": null,
    "decode_tok_s_p50": 72.75199823363032,
    "decode_tok_s_p95": 91.25657936316465,
    "e2e_p50_ms": 873.7625000067055,
    "e2e_p95_ms": 1291.0139000741765,
    "effective_prefill_tok_s_p50": 12506.451393485248,
    "effective_prefill_tok_s_p95": 27012.377256807977,
    "generation_p50_ms": 569.8063999880105,
    "generation_p95_ms": 953.5667999880388,
    "mean_inter_token_latency_ms_p50": 11.870966666416885,
    "mean_inter_token_latency_ms_p95": 15.632242622754735,
    "output_token_sources": {
      "usage": 8
    },
    "output_tokens": 402,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 8,
    "prompt_tokens": 28904,
    "prompt_tokens_p50": 3613.0,
    "prompt_tokens_p95": 3613.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 203.62248451911063,
    "time_to_first_output_p50_ms": 245.17329991795123,
    "time_to_first_output_p95_ms": 337.8761000931263,
    "ttft_p50_ms": 245.17349991947412,
    "ttft_p95_ms": 337.8765000961721,
    "ttft_samples": 8
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k4",
  "request_timings": [
    {
      "content_chunks": 18,
      "decode_tok_s": 72.75199823363032,
      "e2e_ms": 1003.6479000700638,
      "effective_prefill_tok_s": 12506.451393485248,
      "generation_ms": 714.756999979727,
      "mean_inter_token_latency_ms": 13.745326922687058,
      "output_token_source": "usage",
      "output_tokens": 53,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 288.8909000903368,
      "ttft_ms": 288.8916999800131,
      "visible_generation_ms": 714.7562000900507
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 84.23913806691182,
      "e2e_ms": 907.6825000811368,
      "effective_prefill_tok_s": 10693.268920187535,
      "generation_ms": 569.8063999880105,
      "mean_inter_token_latency_ms": 11.870966666416885,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 337.8761000931263,
      "ttft_ms": 337.8765000961721,
      "visible_generation_ms": 569.8059999849647
    },
    {
      "content_chunks": 22,
      "decode_tok_s": 63.970347961742334,
      "e2e_ms": 1291.0139000741765,
      "effective_prefill_tok_s": 10706.863384150392,
      "generation_ms": 953.5667999880388,
      "mean_inter_token_latency_ms": 15.632242622754735,
      "output_token_source": "usage",
      "output_tokens": 62,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 337.44710008613765,
      "ttft_ms": 337.44770009070635,
      "visible_generation_ms": 953.5661999834701
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 87.62944454617976,
      "e2e_ms": 873.7625000067055,
      "effective_prefill_tok_s": 10707.942280121059,
      "generation_ms": 536.34939994663,
      "mean_inter_token_latency_ms": 11.411689360566596,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 337.4131000600755,
      "ttft_ms": 337.41320006083697,
      "visible_generation_ms": 536.3492999458686
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 71.46314514838521,
      "e2e_ms": 777.7215000241995,
      "effective_prefill_tok_s": 17710.853766238037,
      "generation_ms": 573.7223000032827,
      "mean_inter_token_latency_ms": 13.993226829348359,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 203.99920002091676,
      "ttft_ms": 203.9995000232011,
      "visible_generation_ms": 573.7220000009984
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 69.52725953629589,
      "e2e_ms": 935.5499999364838,
      "effective_prefill_tok_s": 14736.51495170605,
      "generation_ms": 690.3767000185326,
      "mean_inter_token_latency_ms": 14.382847917052763,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 245.17329991795123,
      "ttft_ms": 245.17349991947412,
      "visible_generation_ms": 690.3765000170097
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 84.5872340899894,
      "e2e_ms": 713.5945999762043,
      "effective_prefill_tok_s": 22873.574287149913,
      "generation_ms": 555.6394000304863,
      "mean_inter_token_latency_ms": 11.822114894265665,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 157.95519994571805,
      "ttft_ms": 157.95539994724095,
      "visible_generation_ms": 555.6392000289634
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 91.25657936316465,
      "e2e_ms": 681.6591999959201,
      "effective_prefill_tok_s": 27012.377256807977,
      "generation_ms": 547.9057000484318,
      "mean_inter_token_latency_ms": 10.958114000968635,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 133.7534999474883,
      "ttft_ms": 133.7536999490112,
      "visible_generation_ms": 547.9055000469089
    }
  ],
  "requests": 8,
  "run_id": "benchmark-20260905T030924Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": true,
    "reasoning_effort": null,
    "shared_prefix_burst": false,
    "thinking_mode": "disabled"
  },
  "source_recipe": null,
  "started_at": "2026-09-05T03:09:24Z",
  "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": 1974.2417000234127
}