{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 5,
  "concurrency": 1,
  "context_distribution": {
    "131072": 5
  },
  "context_seed": 0,
  "context_tokens": 131072,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-k5",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T23:30:10Z",
  "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": 96,
    "content_chunks_s": null,
    "decode_tok_s_p50": 67.36449093711252,
    "decode_tok_s_p95": 77.62436328628206,
    "e2e_p50_ms": 19892.337500001304,
    "e2e_p95_ms": 25405.160200003593,
    "effective_prefill_tok_s_p50": 5511.197783152851,
    "effective_prefill_tok_s_p95": 5625.363583789455,
    "generation_p50_ms": 721.1361000008765,
    "generation_p95_ms": 754.0087000015774,
    "mean_inter_token_latency_ms_p50": 14.84461600004579,
    "mean_inter_token_latency_ms_p95": 17.16990714287801,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 233,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 528328,
    "prompt_tokens_p50": 105666.0,
    "prompt_tokens_p95": 105668.0,
    "reasoning_chunks": 5,
    "throughput_tok_s": 2.227535848057141,
    "time_to_first_output_p50_ms": 19172.96460000216,
    "time_to_first_output_p95_ms": 24825.445300004503,
    "ttft_p50_ms": 19172.96509999869,
    "ttft_p95_ms": 24947.47990000178,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 13,
      "decode_tok_s": 77.62436328628206,
      "e2e_ms": 25405.160200003593,
      "effective_prefill_tok_s": 4256.237852860622,
      "generation_ms": 579.7148999990895,
      "mean_inter_token_latency_ms": 12.882553333313101,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105663,
      "reasoning_chunks": 5,
      "request_index": 0,
      "time_to_first_output_ms": 24825.445300004503,
      "ttft_ms": 24947.47990000178,
      "visible_generation_ms": 457.6803000018117
    },
    {
      "content_chunks": 22,
      "decode_tok_s": 58.35476434145647,
      "e2e_ms": 19537.50569999829,
      "effective_prefill_tok_s": 5625.363583789455,
      "generation_ms": 754.0087000015774,
      "mean_inter_token_latency_ms": 17.136561363672214,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105664,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 18783.496999996714,
      "ttft_ms": 18783.499299999676,
      "visible_generation_ms": 754.006399998616
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 77.38752875276946,
      "e2e_ms": 19780.297600002086,
      "effective_prefill_tok_s": 5511.197783152851,
      "generation_ms": 607.332999999926,
      "mean_inter_token_latency_ms": 12.92197872340268,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105666,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 19172.96460000216,
      "ttft_ms": 19172.96509999869,
      "visible_generation_ms": 607.3325000033947
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 58.241433205117524,
      "e2e_ms": 19892.337500001304,
      "effective_prefill_tok_s": 5511.808978231153,
      "generation_ms": 721.1361000008765,
      "mean_inter_token_latency_ms": 17.16990714287801,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105668,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 19171.201400000427,
      "ttft_ms": 19171.20189999696,
      "visible_generation_ms": 721.1356000043452
    },
    {
      "content_chunks": 22,
      "decode_tok_s": 67.36449093711252,
      "e2e_ms": 19974.648100003833,
      "effective_prefill_tok_s": 5494.21314813045,
      "generation_ms": 742.2308000022895,
      "mean_inter_token_latency_ms": 14.84461600004579,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105667,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 19232.417300001543,
      "ttft_ms": 19232.41770000459,
      "visible_generation_ms": 742.2303999992437
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260829T232826Z",
  "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": "2026-08-29T23:28:26Z",
  "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": 104599.88789999625
}