{
  "base_url": "http://127.0.0.1:39268/v1",
  "completed": 5,
  "concurrency": 1,
  "context_distribution": {
    "4096": 5
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang-5f55db35-dflash2-k12",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:14:09Z",
  "gpu": "RTX PRO 6000 Blackwell Max-Q TP1",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 512,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 75,
    "content_chunks_s": null,
    "decode_tok_s_p50": 98.22139469855328,
    "decode_tok_s_p95": 114.10181555695192,
    "e2e_p50_ms": 599.4257000274956,
    "e2e_p95_ms": 937.7385000698268,
    "effective_prefill_tok_s_p50": 32866.19011172955,
    "effective_prefill_tok_s_p95": 34062.31528156274,
    "generation_p50_ms": 488.6919000418857,
    "generation_p95_ms": 492.4808000214398,
    "mean_inter_token_latency_ms_p50": 10.181081250872618,
    "mean_inter_token_latency_ms_p95": 10.260016667113328,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 238,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 18065,
    "prompt_tokens_p50": 3613.0,
    "prompt_tokens_p95": 3613.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 74.96595191175648,
    "time_to_first_output_p50_ms": 109.93060003966093,
    "time_to_first_output_p95_ms": 449.04660002794117,
    "ttft_p50_ms": 109.93090004194528,
    "ttft_p95_ms": 449.04720003250986,
    "ttft_samples": 5
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k12",
  "request_timings": [
    {
      "content_chunks": 16,
      "decode_tok_s": 98.22139469855328,
      "e2e_ms": 937.7385000698268,
      "effective_prefill_tok_s": 8045.935543828162,
      "generation_ms": 488.6919000418857,
      "mean_inter_token_latency_ms": 10.181081250872618,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 449.04660002794117,
      "ttft_ms": 449.04720003250986,
      "visible_generation_ms": 488.691300037317
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 97.48388012221695,
      "e2e_ms": 602.3197000613436,
      "effective_prefill_tok_s": 32866.19011172955,
      "generation_ms": 492.3891000216827,
      "mean_inter_token_latency_ms": 10.258106250451723,
      "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": 109.93060003966093,
      "ttft_ms": 109.93090004194528,
      "visible_generation_ms": 492.38880001939833
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 105.31847320592173,
      "e2e_ms": 568.6116999713704,
      "effective_prefill_tok_s": 32015.609923653126,
      "generation_ms": 455.7604999281466,
      "mean_inter_token_latency_ms": 9.495010415169721,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 112.8512000432238,
      "ttft_ms": 112.8516000462696,
      "visible_generation_ms": 455.7600999251008
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 97.46572860893329,
      "e2e_ms": 599.4257000274956,
      "effective_prefill_tok_s": 33783.75219197373,
      "generation_ms": 492.4808000214398,
      "mean_inter_token_latency_ms": 10.260016667113328,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 106.94490000605583,
      "ttft_ms": 106.94510000757873,
      "visible_generation_ms": 492.4806000199169
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 114.10181555695192,
      "e2e_ms": 465.39849997498095,
      "effective_prefill_tok_s": 34062.31528156274,
      "generation_ms": 359.3281999928877,
      "mean_inter_token_latency_ms": 8.76410243885092,
      "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": 106.07029998209327,
      "ttft_ms": 106.07059998437762,
      "visible_generation_ms": 359.3278999906033
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260905T031405Z",
  "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": "2026-09-05T03:14:05Z",
  "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": 3174.774600076489
}