{
  "base_url": "http://127.0.0.1:39062/v1",
  "completed": 16,
  "concurrency": 16,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-r16-b12x-dspark5-sm120",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-02T10:47:44Z",
  "gpu": "dual-rtx-pro-6000-blackwell-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 131072,
  "max_tokens": 512,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 104,
    "content_chunks_s": null,
    "decode_tok_s_p50": 69.07070268176054,
    "decode_tok_s_p95": 87.1214753399235,
    "e2e_p50_ms": 3917.8771000006236,
    "e2e_p95_ms": 6187.771200000498,
    "effective_prefill_tok_s_p50": 2311.005429458654,
    "effective_prefill_tok_s_p95": 3946.4014684881754,
    "generation_p50_ms": 2639.262599999711,
    "generation_p95_ms": 4820.855000001757,
    "mean_inter_token_latency_ms_p50": 14.417249009903292,
    "mean_inter_token_latency_ms_p95": 22.351281132065402,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 3179,
    "prefix_cache_hit_avg": 0.8918918918918919,
    "prompt_token_samples": 16,
    "prompt_tokens": 47360,
    "prompt_tokens_p50": 2960.0,
    "prompt_tokens_p95": 2960.0,
    "reasoning_chunks": 783,
    "throughput_tok_s": 513.5396447759842,
    "time_to_first_output_p50_ms": 1280.3960000001098,
    "time_to_first_output_p95_ms": 1368.489700000282,
    "ttft_p50_ms": 3627.5930000010703,
    "ttft_p95_ms": 6036.375300000145,
    "ttft_samples": 16
  },
  "model": "deepseek-v4-flash-0731-r16-b12x-dspark5-maxseq16-tp2-128k",
  "request_timings": [
    {
      "content_chunks": 5,
      "decode_tok_s": 64.78360509071629,
      "e2e_ms": 3575.8383000011236,
      "effective_prefill_tok_s": 2162.9684169339307,
      "generation_ms": 2207.3486000008415,
      "mean_inter_token_latency_ms": 15.436004195810082,
      "output_token_source": "usage",
      "output_tokens": 144,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 36,
      "request_index": 0,
      "time_to_first_output_ms": 1368.489700000282,
      "ttft_ms": 3352.2910000010597,
      "visible_generation_ms": 223.54730000006384
    },
    {
      "content_chunks": 7,
      "decode_tok_s": 44.74016474005996,
      "e2e_ms": 3119.2861999988963,
      "effective_prefill_tok_s": 3946.4014684881754,
      "generation_ms": 2369.2357999989326,
      "mean_inter_token_latency_ms": 22.351281132065402,
      "output_token_source": "usage",
      "output_tokens": 107,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 26,
      "request_index": 1,
      "time_to_first_output_ms": 750.0503999999637,
      "ttft_ms": 2702.7799999996205,
      "visible_generation_ms": 416.50619999927585
    },
    {
      "content_chunks": 5,
      "decode_tok_s": 67.53267854495633,
      "e2e_ms": 4359.476799998447,
      "effective_prefill_tok_s": 2163.21722435618,
      "generation_ms": 2991.144499999791,
      "mean_inter_token_latency_ms": 14.807646039602927,
      "output_token_source": "usage",
      "output_tokens": 203,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 51,
      "request_index": 2,
      "time_to_first_output_ms": 1368.3322999986558,
      "ttft_ms": 4150.584600000002,
      "visible_generation_ms": 208.892199998445
    },
    {
      "content_chunks": 5,
      "decode_tok_s": 69.3613600842362,
      "e2e_ms": 4279.902600001151,
      "effective_prefill_tok_s": 2164.3465870546743,
      "generation_ms": 2912.284300000465,
      "mean_inter_token_latency_ms": 14.417249009903292,
      "output_token_source": "usage",
      "output_tokens": 203,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 49,
      "request_index": 3,
      "time_to_first_output_ms": 1367.6183000006858,
      "ttft_ms": 4054.70730000161,
      "visible_generation_ms": 225.19529999954102
    },
    {
      "content_chunks": 7,
      "decode_tok_s": 87.1214753399235,
      "e2e_ms": 6187.771200000498,
      "effective_prefill_tok_s": 2165.458277546734,
      "generation_ms": 4820.855000001757,
      "mean_inter_token_latency_ms": 11.478226190480374,
      "output_token_source": "usage",
      "output_tokens": 421,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 109,
      "request_index": 4,
      "time_to_first_output_ms": 1366.916199998741,
      "ttft_ms": 6036.375300000145,
      "visible_generation_ms": 151.3959000003524
    },
    {
      "content_chunks": 8,
      "decode_tok_s": 76.8252459577644,
      "e2e_ms": 5405.524799998602,
      "effective_prefill_tok_s": 2201.791514445193,
      "generation_ms": 4061.164999999164,
      "mean_inter_token_latency_ms": 13.016554487176808,
      "output_token_source": "usage",
      "output_tokens": 313,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 80,
      "request_index": 5,
      "time_to_first_output_ms": 1344.359799999438,
      "ttft_ms": 5166.739199999938,
      "visible_generation_ms": 238.7855999986641
    },
    {
      "content_chunks": 6,
      "decode_tok_s": 69.45396644112223,
      "e2e_ms": 4035.840199998347,
      "effective_prefill_tok_s": 2276.546045874509,
      "generation_ms": 2735.6248999985837,
      "mean_inter_token_latency_ms": 14.39802578946623,
      "output_token_source": "usage",
      "output_tokens": 191,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 46,
      "request_index": 6,
      "time_to_first_output_ms": 1300.2152999997634,
      "ttft_ms": 3794.2917000000307,
      "visible_generation_ms": 241.5484999983164
    },
    {
      "content_chunks": 8,
      "decode_tok_s": 62.872266874926204,
      "e2e_ms": 4018.8240999996196,
      "effective_prefill_tok_s": 2306.879324572022,
      "generation_ms": 2735.70540000037,
      "mean_inter_token_latency_ms": 15.905263953490524,
      "output_token_source": "usage",
      "output_tokens": 173,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 44,
      "request_index": 7,
      "time_to_first_output_ms": 1283.1186999992497,
      "ttft_ms": 3680.351699998937,
      "visible_generation_ms": 338.4724000006827
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 67.02261389751979,
      "e2e_ms": 3205.55129999957,
      "effective_prefill_tok_s": 2311.005429458654,
      "generation_ms": 1924.7235000002547,
      "mean_inter_token_latency_ms": 14.9203372093043,
      "output_token_source": "usage",
      "output_tokens": 130,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 26,
      "request_index": 8,
      "time_to_first_output_ms": 1280.8277999993152,
      "ttft_ms": 2673.4026999984053,
      "visible_generation_ms": 532.1486000011646
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 75.0281047441654,
      "e2e_ms": 5745.389499999874,
      "effective_prefill_tok_s": 2311.7847915799066,
      "generation_ms": 4464.993499999764,
      "mean_inter_token_latency_ms": 13.328338805969445,
      "output_token_source": "usage",
      "output_tokens": 336,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 92,
      "request_index": 9,
      "time_to_first_output_ms": 1280.3960000001098,
      "ttft_ms": 5467.659100000674,
      "visible_generation_ms": 277.7303999992
    },
    {
      "content_chunks": 5,
      "decode_tok_s": 67.31901734847234,
      "e2e_ms": 3433.6325000003853,
      "effective_prefill_tok_s": 2313.025490478491,
      "generation_ms": 2153.923300000315,
      "mean_inter_token_latency_ms": 14.854643448278035,
      "output_token_source": "usage",
      "output_tokens": 146,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 35,
      "request_index": 10,
      "time_to_first_output_ms": 1279.7092000000703,
      "ttft_ms": 3205.308000000514,
      "visible_generation_ms": 228.32449999987148
    },
    {
      "content_chunks": 5,
      "decode_tok_s": 71.57762787805736,
      "e2e_ms": 3263.3511000003637,
      "effective_prefill_tok_s": 2313.419945901582,
      "generation_ms": 1983.8601000010385,
      "mean_inter_token_latency_ms": 13.970845774655201,
      "output_token_source": "usage",
      "output_tokens": 143,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 32,
      "request_index": 11,
      "time_to_first_output_ms": 1279.4909999993251,
      "ttft_ms": 3030.1149999995687,
      "visible_generation_ms": 233.23610000079498
    },
    {
      "content_chunks": 4,
      "decode_tok_s": 70.02774432534929,
      "e2e_ms": 3378.1440000002476,
      "effective_prefill_tok_s": 2314.3514811856676,
      "generation_ms": 2099.1680000006454,
      "mean_inter_token_latency_ms": 14.280054421773098,
      "output_token_source": "usage",
      "output_tokens": 148,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 34,
      "request_index": 12,
      "time_to_first_output_ms": 1278.9759999996022,
      "ttft_ms": 3146.1998999984644,
      "visible_generation_ms": 231.94410000178323
    },
    {
      "content_chunks": 5,
      "decode_tok_s": 69.07070268176054,
      "e2e_ms": 3320.5148999986704,
      "effective_prefill_tok_s": 2314.0757409506928,
      "generation_ms": 2041.3864999991347,
      "mean_inter_token_latency_ms": 14.477918439710175,
      "output_token_source": "usage",
      "output_tokens": 142,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 33,
      "request_index": 13,
      "time_to_first_output_ms": 1279.1283999995358,
      "ttft_ms": 3089.7822999995697,
      "visible_generation_ms": 230.73259999910078
    },
    {
      "content_chunks": 8,
      "decode_tok_s": 66.97007599117197,
      "e2e_ms": 4145.994000000428,
      "effective_prefill_tok_s": 2314.2322389886367,
      "generation_ms": 2866.952099999253,
      "mean_inter_token_latency_ms": 14.932042187496108,
      "output_token_source": "usage",
      "output_tokens": 193,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 47,
      "request_index": 14,
      "time_to_first_output_ms": 1279.0419000011752,
      "ttft_ms": 3821.1547000009887,
      "visible_generation_ms": 324.8392999994394
    },
    {
      "content_chunks": 7,
      "decode_tok_s": 70.09533647770414,
      "e2e_ms": 3917.8771000006236,
      "effective_prefill_tok_s": 2315.0058129310178,
      "generation_ms": 2639.262599999711,
      "mean_inter_token_latency_ms": 14.266284324322761,
      "output_token_source": "usage",
      "output_tokens": 186,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2960,
      "reasoning_chunks": 43,
      "request_index": 15,
      "time_to_first_output_ms": 1278.6145000009128,
      "ttft_ms": 3627.5930000010703,
      "visible_generation_ms": 290.2840999995533
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260802T104738Z",
  "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": "2026-08-02T10:47:38Z",
  "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": 6190.369199999623
}
