{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 12,
  "concurrency": 4,
  "context_distribution": {
    "4096": 12
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-k5",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T23:45:20Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 1048576,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 171,
    "content_chunks_s": null,
    "decode_tok_s_p50": 15.076313961852573,
    "decode_tok_s_p95": 63.94474303456862,
    "e2e_p50_ms": 4278.277999997954,
    "e2e_p95_ms": 6268.716500002483,
    "effective_prefill_tok_s_p50": 1529.7291889645908,
    "effective_prefill_tok_s_p95": 2891.506919347136,
    "generation_p50_ms": 1923.5470999992685,
    "generation_p95_ms": 3476.472399997874,
    "mean_inter_token_latency_ms_p50": 57.26170408161124,
    "mean_inter_token_latency_ms_p95": 87.20082571422348,
    "output_token_sources": {
      "usage": 12
    },
    "output_tokens": 496,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 12,
    "prompt_tokens": 35616,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 12,
    "throughput_tok_s": 37.217825012072545,
    "time_to_first_output_p50_ms": 1927.8889000052004,
    "time_to_first_output_p95_ms": 4019.221700000344,
    "ttft_p50_ms": 1940.3355000031297,
    "ttft_p95_ms": 4019.2222999976366,
    "ttft_samples": 12
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 15.076313961852573,
      "e2e_ms": 5942.768799999612,
      "effective_prefill_tok_s": 738.4514270510995,
      "generation_ms": 1923.5470999992685,
      "mean_inter_token_latency_ms": 66.32921034480236,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 4019.221700000344,
      "ttft_ms": 4019.2222999976366,
      "visible_generation_ms": 1923.5465000019758
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 12.794040598287072,
      "e2e_ms": 6268.716500002483,
      "effective_prefill_tok_s": 968.6369519007307,
      "generation_ms": 3204.6170000030543,
      "mean_inter_token_latency_ms": 78.16139024397694,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 3064.099499999429,
      "ttft_ms": 3064.1011000043363,
      "visible_generation_ms": 3204.615399998147
    },
    {
      "content_chunks": 9,
      "decode_tok_s": 11.467781317544201,
      "e2e_ms": 4257.155300001614,
      "effective_prefill_tok_s": 2462.8121996088225,
      "generation_ms": 3052.028899997822,
      "mean_inter_token_latency_ms": 87.20082571422348,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 1205.1264000037918,
      "ttft_ms": 2434.529600002861,
      "visible_generation_ms": 1822.6256999987527
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 22.632715512258798,
      "e2e_ms": 4959.17599999666,
      "effective_prefill_tok_s": 806.9943132462998,
      "generation_ms": 1281.3309999983176,
      "mean_inter_token_latency_ms": 44.18382758614888,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 3677.8449999983422,
      "ttft_ms": 3677.8456999963964,
      "visible_generation_ms": 1281.3303000002634
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 12.80052311054807,
      "e2e_ms": 4278.277999997954,
      "effective_prefill_tok_s": 2760.2012826620376,
      "generation_ms": 3202.994100000069,
      "mean_inter_token_latency_ms": 78.12180731707485,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 1075.2838999978849,
      "ttft_ms": 1075.284599995939,
      "visible_generation_ms": 3202.993400002015
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 11.630559049374167,
      "e2e_ms": 4704.737799998838,
      "effective_prefill_tok_s": 2196.074206586801,
      "generation_ms": 3353.2351999965613,
      "mean_inter_token_latency_ms": 85.98038974350158,
      "output_token_source": "usage",
      "output_tokens": 40,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 1351.5026000022772,
      "ttft_ms": 1351.5030999988085,
      "visible_generation_ms": 3353.23470000003
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 24.350499777533983,
      "e2e_ms": 3721.8616000027396,
      "effective_prefill_tok_s": 1347.6295885688985,
      "generation_ms": 1519.4760000013048,
      "mean_inter_token_latency_ms": 41.06691891895419,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 2202.385600001435,
      "ttft_ms": 2202.386299999489,
      "visible_generation_ms": 1519.4753000032506
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 17.463678666893454,
      "e2e_ms": 4733.712400004151,
      "effective_prefill_tok_s": 1539.507800471279,
      "generation_ms": 2805.823499998951,
      "mean_inter_token_latency_ms": 57.26170408161124,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 1927.8889000052004,
      "ttft_ms": 1927.8892000002088,
      "visible_generation_ms": 2805.8232000039425
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 14.382395211890817,
      "e2e_ms": 4502.926799999841,
      "effective_prefill_tok_s": 2891.506919347136,
      "generation_ms": 3476.472399997874,
      "mean_inter_token_latency_ms": 69.52944799995748,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 8,
      "time_to_first_output_ms": 1026.4544000019669,
      "ttft_ms": 1128.1071000048541,
      "visible_generation_ms": 3374.819699994987
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 28.584738900663634,
      "e2e_ms": 3374.6669000029215,
      "effective_prefill_tok_s": 1529.632690765565,
      "generation_ms": 1434.3318000028376,
      "mean_inter_token_latency_ms": 34.9837024390936,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 1940.335100000084,
      "ttft_ms": 1940.3355000031297,
      "visible_generation_ms": 1434.3313999997918
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 28.62385467257793,
      "e2e_ms": 3547.2638999999617,
      "effective_prefill_tok_s": 1529.7291889645908,
      "generation_ms": 1607.0511999932933,
      "mean_inter_token_latency_ms": 34.935895652028115,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 10,
      "time_to_first_output_ms": 1940.2127000066685,
      "ttft_ms": 2598.244300002989,
      "visible_generation_ms": 949.0195999969728
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 63.94474303456862,
      "e2e_ms": 2322.707100000116,
      "effective_prefill_tok_s": 1869.3737314579798,
      "generation_ms": 735.0096000009216,
      "mean_inter_token_latency_ms": 15.638502127679184,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 11,
      "time_to_first_output_ms": 1587.6974999991944,
      "ttft_ms": 1688.1990999972913,
      "visible_generation_ms": 634.5080000028247
    }
  ],
  "requests": 12,
  "run_id": "benchmark-20260829T234507Z",
  "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": "2026-08-29T23:45:07Z",
  "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": 13326.947499997914
}