{
  "base_url": "http://127.0.0.1:39051/v1",
  "completed": 12,
  "concurrency": 1,
  "context_distribution": {
    "30720": 12
  },
  "context_seed": 0,
  "context_tokens": 32768,
  "engine": "sglang-b7252cc6b-sm120-compat",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-01T16:01:42Z",
  "gpu": "2x RTX PRO 6000 Blackwell Max-Q TP2",
  "manifest": null,
  "max_context_tokens": 32768,
  "max_tokens": 1024,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 469,
    "content_chunks_s": null,
    "decode_tok_s_p50": 73.53745963993185,
    "decode_tok_s_p95": 74.04758800030926,
    "e2e_p50_ms": 5179.705799997464,
    "e2e_p95_ms": 6138.166500000807,
    "effective_prefill_tok_s_p50": 7843.75979058298,
    "effective_prefill_tok_s_p95": 7881.604972988148,
    "generation_p50_ms": 2390.3546999972605,
    "generation_p95_ms": 3343.528699999297,
    "mean_inter_token_latency_ms_p50": 13.591580081297954,
    "mean_inter_token_latency_ms_p95": 13.69526048385465,
    "output_token_sources": {
      "usage": 12
    },
    "output_tokens": 2171,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 12,
    "prompt_tokens": 262552,
    "prompt_tokens_p50": 21879.0,
    "prompt_tokens_p95": 21882.0,
    "reasoning_chunks": 1618,
    "throughput_tok_s": 34.55035332287966,
    "time_to_first_output_p50_ms": 2788.6455000007118,
    "time_to_first_output_p95_ms": 2803.9595000009285,
    "ttft_p50_ms": 4630.146999999852,
    "ttft_p95_ms": 5425.789000000805,
    "ttft_samples": 12
  },
  "model": "inkling-small-nvfp4-tp2",
  "request_timings": [
    {
      "content_chunks": 33,
      "decode_tok_s": 73.01796129974312,
      "e2e_ms": 4502.171799998905,
      "effective_prefill_tok_s": 7803.9643582558,
      "generation_ms": 1698.2122999979765,
      "mean_inter_token_latency_ms": 13.69526048385465,
      "output_token_source": "usage",
      "output_tokens": 125,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21882,
      "reasoning_chunks": 85,
      "request_index": 0,
      "time_to_first_output_ms": 2803.9595000009285,
      "ttft_ms": 4036.107500000071,
      "visible_generation_ms": 466.06429999883403
    },
    {
      "content_chunks": 34,
      "decode_tok_s": 73.15210524993282,
      "e2e_ms": 5359.868899999128,
      "effective_prefill_tok_s": 7881.604972988148,
      "generation_ms": 2583.657699997275,
      "mean_inter_token_latency_ms": 13.670146560832142,
      "output_token_source": "usage",
      "output_tokens": 190,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21881,
      "reasoning_chunks": 149,
      "request_index": 1,
      "time_to_first_output_ms": 2776.2112000018533,
      "ttft_ms": 4878.44799999948,
      "visible_generation_ms": 481.4208999996481
    },
    {
      "content_chunks": 32,
      "decode_tok_s": 73.93003820327213,
      "e2e_ms": 4392.033200001606,
      "effective_prefill_tok_s": 7864.064853565131,
      "generation_ms": 1609.6298999982537,
      "mean_inter_token_latency_ms": 13.526301680657594,
      "output_token_source": "usage",
      "output_tokens": 120,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21881,
      "reasoning_chunks": 81,
      "request_index": 2,
      "time_to_first_output_ms": 2782.4033000033523,
      "ttft_ms": 3938.568200002919,
      "visible_generation_ms": 453.46499999868684
    },
    {
      "content_chunks": 39,
      "decode_tok_s": 73.21234107244473,
      "e2e_ms": 4987.728299998707,
      "effective_prefill_tok_s": 7845.744466263071,
      "generation_ms": 2199.0827999979956,
      "mean_inter_token_latency_ms": 13.658899378869538,
      "output_token_source": "usage",
      "output_tokens": 162,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21879,
      "reasoning_chunks": 116,
      "request_index": 3,
      "time_to_first_output_ms": 2788.6455000007118,
      "ttft_ms": 4439.525599998888,
      "visible_generation_ms": 548.2026999998197
    },
    {
      "content_chunks": 49,
      "decode_tok_s": 73.49046048366122,
      "e2e_ms": 5424.004200001946,
      "effective_prefill_tok_s": 7858.255305689428,
      "generation_ms": 2639.798399999563,
      "mean_inter_token_latency_ms": 13.607208247420429,
      "output_token_source": "usage",
      "output_tokens": 195,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21879,
      "reasoning_chunks": 139,
      "request_index": 4,
      "time_to_first_output_ms": 2784.2058000023826,
      "ttft_ms": 4738.3401000006415,
      "visible_generation_ms": 685.6641000013042
    },
    {
      "content_chunks": 32,
      "decode_tok_s": 73.71247344053347,
      "e2e_ms": 4792.9303999990225,
      "effective_prefill_tok_s": 7855.650210252386,
      "generation_ms": 2007.8012999983912,
      "mean_inter_token_latency_ms": 13.56622499998913,
      "output_token_source": "usage",
      "output_tokens": 149,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21879,
      "reasoning_chunks": 110,
      "request_index": 5,
      "time_to_first_output_ms": 2785.1291000006313,
      "ttft_ms": 4342.105499999889,
      "visible_generation_ms": 450.8248999991338
    },
    {
      "content_chunks": 47,
      "decode_tok_s": 73.45558574641552,
      "e2e_ms": 5863.648799997463,
      "effective_prefill_tok_s": 7850.490267228458,
      "generation_ms": 3076.6890999984753,
      "mean_inter_token_latency_ms": 13.61366858406405,
      "output_token_source": "usage",
      "output_tokens": 227,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21879,
      "reasoning_chunks": 173,
      "request_index": 6,
      "time_to_first_output_ms": 2786.959699998988,
      "ttft_ms": 5205.025899998873,
      "visible_generation_ms": 658.6228999985906
    },
    {
      "content_chunks": 51,
      "decode_tok_s": 73.57496288279258,
      "e2e_ms": 6138.166500000807,
      "effective_prefill_tok_s": 7828.563687211336,
      "generation_ms": 3343.528699999297,
      "mean_inter_token_latency_ms": 13.591580081297954,
      "output_token_source": "usage",
      "output_tokens": 247,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21878,
      "reasoning_chunks": 189,
      "request_index": 7,
      "time_to_first_output_ms": 2794.63780000151,
      "ttft_ms": 5425.789000000805,
      "visible_generation_ms": 712.3775000000023
    },
    {
      "content_chunks": 39,
      "decode_tok_s": 74.04758800030926,
      "e2e_ms": 5179.705799997464,
      "effective_prefill_tok_s": 7843.75979058298,
      "generation_ms": 2390.3546999972605,
      "mean_inter_token_latency_ms": 13.504828813543844,
      "output_token_source": "usage",
      "output_tokens": 178,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21879,
      "reasoning_chunks": 132,
      "request_index": 8,
      "time_to_first_output_ms": 2789.3511000002036,
      "ttft_ms": 4630.146999999852,
      "visible_generation_ms": 549.5587999976124
    },
    {
      "content_chunks": 39,
      "decode_tok_s": 73.66097296511911,
      "e2e_ms": 5536.515100000543,
      "effective_prefill_tok_s": 7830.087354920477,
      "generation_ms": 2742.2934000023815,
      "mean_inter_token_latency_ms": 13.575709901001888,
      "output_token_source": "usage",
      "output_tokens": 203,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21879,
      "reasoning_chunks": 157,
      "request_index": 9,
      "time_to_first_output_ms": 2794.221699998161,
      "ttft_ms": 4986.724200000026,
      "visible_generation_ms": 549.7909000005166
    },
    {
      "content_chunks": 34,
      "decode_tok_s": 73.7347015167875,
      "e2e_ms": 5519.826399999147,
      "effective_prefill_tok_s": 7830.807034858056,
      "generation_ms": 2725.9892000001855,
      "mean_inter_token_latency_ms": 13.562135323384007,
      "output_token_source": "usage",
      "output_tokens": 202,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21878,
      "reasoning_chunks": 161,
      "request_index": 10,
      "time_to_first_output_ms": 2793.8371999989613,
      "ttft_ms": 5039.27819999808,
      "visible_generation_ms": 480.548200001067
    },
    {
      "content_chunks": 40,
      "decode_tok_s": 73.53745963993185,
      "e2e_ms": 5133.510299998306,
      "effective_prefill_tok_s": 7828.763983880715,
      "generation_ms": 2338.9439999991737,
      "mean_inter_token_latency_ms": 13.598511627902173,
      "output_token_source": "usage",
      "output_tokens": 173,
      "planned_context_tokens": 30720,
      "prompt_tokens": 21878,
      "reasoning_chunks": 126,
      "request_index": 11,
      "time_to_first_output_ms": 2794.5662999991328,
      "ttft_ms": 4562.683699998161,
      "visible_generation_ms": 570.8266000001458
    }
  ],
  "requests": 12,
  "run_id": "benchmark-20260801T160039Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": "low",
    "shared_prefix_burst": false,
    "thinking_mode": "default"
  },
  "source_recipe": null,
  "started_at": "2026-08-01T16:00:39Z",
  "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": 62835.82629999728
}