{
  "base_url": "http://127.0.0.1:8081/v1",
  "completed": 1,
  "concurrency": 1,
  "context_distribution": {
    "252928": 1
  },
  "context_seed": 0,
  "context_tokens": 252928,
  "engine": "sglang",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-03T18:56:35Z",
  "gpu": "rtx-5090",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 8192,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 72,
    "content_chunks_s": null,
    "decode_tok_s_p50": 57.373370596495725,
    "decode_tok_s_p95": 57.373370596495725,
    "e2e_p50_ms": 97950.17790002748,
    "e2e_p95_ms": 97950.17790002748,
    "effective_prefill_tok_s_p50": 2523.4127268611633,
    "effective_prefill_tok_s_p95": 2523.4127268611633,
    "generation_p50_ms": 1254.9375999951735,
    "generation_p95_ms": 1254.9375999951735,
    "mean_inter_token_latency_ms_p50": 17.429688888821854,
    "mean_inter_token_latency_ms_p95": 17.429688888821854,
    "output_token_sources": {
      "usage": 1
    },
    "output_tokens": 73,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 1,
    "prompt_tokens": 244002,
    "prompt_tokens_p50": 244002.0,
    "prompt_tokens_p95": 244002.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 0.7452297210067808,
    "time_to_first_output_p50_ms": 96695.24030003231,
    "time_to_first_output_p95_ms": 96695.24030003231,
    "ttft_p50_ms": 96695.24120003916,
    "ttft_p95_ms": 96695.24120003916,
    "ttft_samples": 1
  },
  "model": "qwen38-27b-gittensor-nvfp4-nospec-262k",
  "request_timings": [
    {
      "content_chunks": 72,
      "decode_tok_s": 57.373370596495725,
      "e2e_ms": 97950.17790002748,
      "effective_prefill_tok_s": 2523.4127268611633,
      "generation_ms": 1254.9375999951735,
      "mean_inter_token_latency_ms": 17.429688888821854,
      "output_token_source": "usage",
      "output_tokens": 73,
      "planned_context_tokens": 252928,
      "prompt_tokens": 244002,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 96695.24030003231,
      "ttft_ms": 96695.24120003916,
      "visible_generation_ms": 1254.9366999883205
    }
  ],
  "requests": 1,
  "run_id": "benchmark-20260903T185457Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": null,
    "shared_prefix_burst": false,
    "thinking_mode": "disabled"
  },
  "source_recipe": null,
  "started_at": "2026-09-03T18:54:57Z",
  "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": 97956.3723000465
}