{
  "base_url": "http://127.0.0.1:39053/v1",
  "completed": 12,
  "concurrency": 1,
  "context_distribution": {
    "32768": 12
  },
  "context_seed": 0,
  "context_tokens": 32768,
  "engine": "vllm",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-01T12:06:23Z",
  "gpu": "dual-rtx-pro-6000-blackwell-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 660,
    "content_chunks_s": null,
    "decode_tok_s_p50": 67.53151173232763,
    "decode_tok_s_p95": 68.50786509073566,
    "e2e_p50_ms": 3141.726399999243,
    "e2e_p95_ms": 3289.3682999983866,
    "effective_prefill_tok_s_p50": 12820.701413283294,
    "effective_prefill_tok_s_p95": 12835.871504515932,
    "generation_p50_ms": 818.2723999998416,
    "generation_p95_ms": 964.4564999998693,
    "mean_inter_token_latency_ms_p50": 14.765828301879425,
    "mean_inter_token_latency_ms_p95": 14.862079310346573,
    "output_token_sources": {
      "usage": 12
    },
    "output_tokens": 672,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 12,
    "prompt_tokens": 357644,
    "prompt_tokens_p50": 29804.0,
    "prompt_tokens_p95": 29814.0,
    "throughput_tok_s": 17.827936062759612,
    "ttft_p50_ms": 2324.429800000871,
    "ttft_p95_ms": 2360.6635999985883,
    "ttft_samples": 12
  },
  "model": "qwen35-122b-a10b-nvfp4-tp2",
  "request_timings": [
    {
      "decode_tok_s": 67.97117924896747,
      "e2e_ms": 3184.5421999987593,
      "effective_prefill_tok_s": 12620.180190018526,
      "generation_ms": 823.878600000171,
      "mean_inter_token_latency_ms": 14.71211785714591,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29792,
      "request_index": 0,
      "ttft_ms": 2360.6635999985883
    },
    {
      "decode_tok_s": 68.43686772279113,
      "e2e_ms": 3141.726399999243,
      "effective_prefill_tok_s": 12823.580755206549,
      "generation_ms": 818.2723999998416,
      "mean_inter_token_latency_ms": 14.612007142854313,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29795,
      "request_index": 1,
      "ttft_ms": 2323.4539999994013
    },
    {
      "decode_tok_s": 68.14476118667076,
      "e2e_ms": 3187.3643999988417,
      "effective_prefill_tok_s": 12834.901351921302,
      "generation_ms": 865.8039000001736,
      "mean_inter_token_latency_ms": 14.674642372884298,
      "output_token_source": "usage",
      "output_tokens": 60,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29797,
      "request_index": 2,
      "ttft_ms": 2321.560499998668
    },
    {
      "decode_tok_s": 67.28533599627795,
      "e2e_ms": 3195.274800000334,
      "effective_prefill_tok_s": 12771.752244119885,
      "generation_ms": 862.0006000001013,
      "mean_inter_token_latency_ms": 14.862079310346573,
      "output_token_source": "usage",
      "output_tokens": 59,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29800,
      "request_index": 3,
      "ttft_ms": 2333.2742000002327
    },
    {
      "decode_tok_s": 68.50786509073566,
      "e2e_ms": 3097.949499999231,
      "effective_prefill_tok_s": 12821.408038505117,
      "generation_ms": 773.6337999995158,
      "mean_inter_token_latency_ms": 14.596864150934259,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29801,
      "request_index": 4,
      "ttft_ms": 2324.3156999997154
    },
    {
      "decode_tok_s": 67.53151173232763,
      "e2e_ms": 3111.2115000014455,
      "effective_prefill_tok_s": 12811.250654281506,
      "generation_ms": 784.8188000007212,
      "mean_inter_token_latency_ms": 14.80790188680606,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29804,
      "request_index": 5,
      "ttft_ms": 2326.3927000007243
    },
    {
      "decode_tok_s": 67.41877554463825,
      "e2e_ms": 3214.26799999972,
      "effective_prefill_tok_s": 12823.600587913807,
      "generation_ms": 889.9598000007245,
      "mean_inter_token_latency_ms": 14.832663333345408,
      "output_token_source": "usage",
      "output_tokens": 61,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29806,
      "request_index": 6,
      "ttft_ms": 2324.3081999989954
    },
    {
      "decode_tok_s": 67.72393526157404,
      "e2e_ms": 3104.7529999996186,
      "effective_prefill_tok_s": 12835.871504515932,
      "generation_ms": 782.5888999996096,
      "mean_inter_token_latency_ms": 14.765828301879425,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29807,
      "request_index": 7,
      "ttft_ms": 2322.164100000009
    },
    {
      "decode_tok_s": 67.39547092067792,
      "e2e_ms": 3289.3682999983866,
      "effective_prefill_tok_s": 12820.701413283294,
      "generation_ms": 964.4564999998693,
      "mean_inter_token_latency_ms": 14.837792307690297,
      "output_token_source": "usage",
      "output_tokens": 66,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29807,
      "request_index": 8,
      "ttft_ms": 2324.9117999985174
    },
    {
      "decode_tok_s": 67.78624187279307,
      "e2e_ms": 2937.0555999994394,
      "effective_prefill_tok_s": 12781.422075031465,
      "generation_ms": 604.8425000008137,
      "mean_inter_token_latency_ms": 14.752256097580823,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29809,
      "request_index": 9,
      "ttft_ms": 2332.2130999986257
    },
    {
      "decode_tok_s": 67.37394848113496,
      "e2e_ms": 3200.1392000001943,
      "effective_prefill_tok_s": 12825.511013491923,
      "generation_ms": 875.7093999993231,
      "mean_inter_token_latency_ms": 14.842532203378356,
      "output_token_source": "usage",
      "output_tokens": 60,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29812,
      "request_index": 10,
      "ttft_ms": 2324.429800000871
    },
    {
      "decode_tok_s": 67.30244084468335,
      "e2e_ms": 3023.8717000011093,
      "effective_prefill_tok_s": 12820.295059427306,
      "generation_ms": 698.3402000005299,
      "mean_inter_token_latency_ms": 14.858302127670848,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 32768,
      "prompt_tokens": 29814,
      "request_index": 11,
      "ttft_ms": 2325.5315000005794
    }
  ],
  "requests": 12,
  "run_id": "benchmark-20260801T120545Z",
  "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": "2026-08-01T12:05:45Z",
  "tier": null,
  "timing_methodology": {
    "clock": "client time.perf_counter",
    "decode": "usage completion tokens after the first token divided by client-observed generation time",
    "effective_prefill": "usage.prompt_tokens divided by client-observed TTFT; includes queueing, scheduling, prefill, and first-token work",
    "generation": "client-observed E2E minus TTFT",
    "mean_inter_token_latency": "client-observed generation time divided by completion tokens after the first token; not a raw per-token timestamp trace",
    "ttft": "request start through the first non-empty streamed content delta"
  },
  "wall_clock_ms": 37693.65100000141
}