{
  "base_url": "http://127.0.0.1:39051/v1",
  "completed": 12,
  "concurrency": 1,
  "context_distribution": {
    "31488": 12
  },
  "context_seed": 0,
  "context_tokens": 32768,
  "engine": "sglang-b7252cc6b-sm120-compat",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-01T16:00:39Z",
  "gpu": "2x RTX PRO 6000 Blackwell Max-Q TP2",
  "manifest": null,
  "max_context_tokens": 32768,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 692,
    "content_chunks_s": null,
    "decode_tok_s_p50": 74.60970645153735,
    "decode_tok_s_p95": 76.1501528714815,
    "e2e_p50_ms": 3567.578500002128,
    "e2e_p95_ms": 4129.924799999571,
    "effective_prefill_tok_s_p50": 7886.529648370295,
    "effective_prefill_tok_s_p95": 7920.014458981735,
    "generation_p50_ms": 721.7080000009446,
    "generation_p95_ms": 1152.8889000001072,
    "mean_inter_token_latency_ms_p50": 13.403029629657402,
    "mean_inter_token_latency_ms_p95": 13.661942666658433,
    "output_token_sources": {
      "usage": 12
    },
    "output_tokens": 740,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 12,
    "prompt_tokens": 269056,
    "prompt_tokens_p50": 22421.0,
    "prompt_tokens_p95": 22424.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 16.776964971358574,
    "time_to_first_output_p50_ms": 2838.3444999999483,
    "time_to_first_output_p95_ms": 3105.279100000189,
    "ttft_p50_ms": 2838.3450000001176,
    "ttft_p95_ms": 3105.280399999174,
    "ttft_samples": 12
  },
  "model": "inkling-small-nvfp4-tp2",
  "request_timings": [
    {
      "content_chunks": 72,
      "decode_tok_s": 73.1960325408531,
      "e2e_ms": 4129.924799999571,
      "effective_prefill_tok_s": 7221.251062424191,
      "generation_ms": 1024.6456999993825,
      "mean_inter_token_latency_ms": 13.661942666658433,
      "output_token_source": "usage",
      "output_tokens": 76,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22424,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 3105.279100000189,
      "ttft_ms": 3105.280399999174,
      "visible_generation_ms": 1024.6444000003976
    },
    {
      "content_chunks": 50,
      "decode_tok_s": 73.4369024590702,
      "e2e_ms": 3557.8825999982655,
      "effective_prefill_tok_s": 7906.071791215245,
      "generation_ms": 721.7080000009446,
      "mean_inter_token_latency_ms": 13.61713207548952,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22423,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 2836.174599997321,
      "ttft_ms": 2836.1753999997745,
      "visible_generation_ms": 721.707199998491
    },
    {
      "content_chunks": 43,
      "decode_tok_s": 75.06968425021677,
      "e2e_ms": 3443.945700000768,
      "effective_prefill_tok_s": 7920.014458981735,
      "generation_ms": 612.7640000013344,
      "mean_inter_token_latency_ms": 13.320956521768139,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22423,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 2831.1816999994335,
      "ttft_ms": 2831.18199999808,
      "visible_generation_ms": 612.763700002688
    },
    {
      "content_chunks": 44,
      "decode_tok_s": 74.24945256819277,
      "e2e_ms": 3468.0336999990686,
      "effective_prefill_tok_s": 7908.55159186619,
      "generation_ms": 633.0013000006147,
      "mean_inter_token_latency_ms": 13.468112765970526,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22421,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 2835.032399998454,
      "ttft_ms": 2835.032899998623,
      "visible_generation_ms": 633.0008000004455
    },
    {
      "content_chunks": 42,
      "decode_tok_s": 75.67876282375319,
      "e2e_ms": 3432.963100000052,
      "effective_prefill_tok_s": 7899.323003250806,
      "generation_ms": 594.6186000001035,
      "mean_inter_token_latency_ms": 13.213746666668966,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22421,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 2838.3444999999483,
      "ttft_ms": 2838.3450000001176,
      "visible_generation_ms": 594.6180999999342
    },
    {
      "content_chunks": 41,
      "decode_tok_s": 76.1501528714815,
      "e2e_ms": 3409.5312999997986,
      "effective_prefill_tok_s": 7917.787229022945,
      "generation_ms": 577.8057999996236,
      "mean_inter_token_latency_ms": 13.131949999991445,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22421,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 2831.725500000175,
      "ttft_ms": 2831.7257999988215,
      "visible_generation_ms": 577.8055000009772
    },
    {
      "content_chunks": 78,
      "decode_tok_s": 74.60999696570553,
      "e2e_ms": 3932.914200002415,
      "effective_prefill_tok_s": 7874.5638627440785,
      "generation_ms": 1085.6454000022495,
      "mean_inter_token_latency_ms": 13.403029629657402,
      "output_token_source": "usage",
      "output_tokens": 82,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22421,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 2847.2688000001654,
      "ttft_ms": 2847.269400001096,
      "visible_generation_ms": 1085.6448000013188
    },
    {
      "content_chunks": 56,
      "decode_tok_s": 75.57794712352724,
      "e2e_ms": 3626.1029000015697,
      "effective_prefill_tok_s": 7879.240552263445,
      "generation_ms": 780.6510000009439,
      "mean_inter_token_latency_ms": 13.23137288137193,
      "output_token_source": "usage",
      "output_tokens": 60,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22420,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 2845.4519000006258,
      "ttft_ms": 2845.45220000291,
      "visible_generation_ms": 780.6506999986595
    },
    {
      "content_chunks": 58,
      "decode_tok_s": 73.72484124671347,
      "e2e_ms": 3670.3497000016796,
      "effective_prefill_tok_s": 7886.529648370295,
      "generation_ms": 827.4009000015212,
      "mean_inter_token_latency_ms": 13.563949180352807,
      "output_token_source": "usage",
      "output_tokens": 62,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22421,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 2842.9488000001584,
      "ttft_ms": 2842.949199999566,
      "visible_generation_ms": 827.4005000021134
    },
    {
      "content_chunks": 83,
      "decode_tok_s": 74.59521901892889,
      "e2e_ms": 3996.3079999979527,
      "effective_prefill_tok_s": 7885.225220586367,
      "generation_ms": 1152.8889000001072,
      "mean_inter_token_latency_ms": 13.405684883722177,
      "output_token_source": "usage",
      "output_tokens": 87,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22421,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 2843.4190999978455,
      "ttft_ms": 2843.4195999980147,
      "visible_generation_ms": 1152.888399999938
    },
    {
      "content_chunks": 74,
      "decode_tok_s": 74.60970645153735,
      "e2e_ms": 3866.441099999065,
      "effective_prefill_tok_s": 7909.95270328224,
      "generation_ms": 1032.0372999995016,
      "mean_inter_token_latency_ms": 13.403081818175345,
      "output_token_source": "usage",
      "output_tokens": 78,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22420,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 2834.4037999995635,
      "ttft_ms": 2834.4042999997328,
      "visible_generation_ms": 1032.0367999993323
    },
    {
      "content_chunks": 51,
      "decode_tok_s": 74.95412183098624,
      "e2e_ms": 3567.578500002128,
      "effective_prefill_tok_s": 7874.574950325253,
      "generation_ms": 720.4407000026549,
      "mean_inter_token_latency_ms": 13.341494444493609,
      "output_token_source": "usage",
      "output_tokens": 55,
      "planned_context_tokens": 31488,
      "prompt_tokens": 22420,
      "reasoning_chunks": 0,
      "request_index": 11,
      "time_to_first_output_ms": 2847.137799999473,
      "ttft_ms": 2847.1381999988807,
      "visible_generation_ms": 720.4403000032471
    }
  ],
  "requests": 12,
  "run_id": "benchmark-20260801T155955Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": "none",
    "shared_prefix_burst": false,
    "thinking_mode": "default"
  },
  "source_recipe": null,
  "started_at": "2026-08-01T15:59:55Z",
  "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": 44108.09710000103
}