{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 5,
  "concurrency": 1,
  "context_distribution": {
    "240000": 5
  },
  "context_seed": 0,
  "context_tokens": 240000,
  "engine": "vllm-glm53-exl3-k3-b12x-dflash2-k5",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T23:34:03Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 1048576,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 86,
    "content_chunks_s": null,
    "decode_tok_s_p50": 67.90368482256623,
    "decode_tok_s_p95": 83.91790656686774,
    "e2e_p50_ms": 38543.67239999556,
    "e2e_p95_ms": 61236.73940000299,
    "effective_prefill_tok_s_p50": 5193.455993296913,
    "effective_prefill_tok_s_p95": 6273.350591707508,
    "generation_p50_ms": 607.7368000042043,
    "generation_p95_ms": 708.6178000026848,
    "mean_inter_token_latency_ms_p50": 14.72674130443762,
    "mean_inter_token_latency_ms_p95": 17.043743478058353,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 212,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 990708,
    "prompt_tokens_p50": 198142.0,
    "prompt_tokens_p95": 198144.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 0.9404185758810596,
    "time_to_first_output_p50_ms": 38151.666300000215,
    "time_to_first_output_p95_ms": 60528.121600000304,
    "ttft_p50_ms": 38151.66709999903,
    "ttft_p95_ms": 60528.12200000335,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 58.67255637163119,
      "e2e_ms": 38543.67239999556,
      "effective_prefill_tok_s": 5193.455993296913,
      "generation_ms": 392.0060999953421,
      "mean_inter_token_latency_ms": 17.043743478058353,
      "output_token_source": "usage",
      "output_tokens": 24,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198139,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 38151.666300000215,
      "ttft_ms": 38151.66709999903,
      "visible_generation_ms": 392.00529999652645
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 83.91790656686774,
      "e2e_ms": 32192.134500000975,
      "effective_prefill_tok_s": 6273.350591707508,
      "generation_ms": 607.7368000042043,
      "mean_inter_token_latency_ms": 11.916407843219691,
      "output_token_source": "usage",
      "output_tokens": 52,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198140,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 31584.39769999677,
      "ttft_ms": 31584.398500002862,
      "visible_generation_ms": 607.7359999981127
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 76.82514204520773,
      "e2e_ms": 32409.30720000324,
      "effective_prefill_tok_s": 6221.1780868325695,
      "generation_ms": 559.7126000066055,
      "mean_inter_token_latency_ms": 13.016572093176872,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198142,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 31849.594599996635,
      "ttft_ms": 31849.595100000442,
      "visible_generation_ms": 559.7121000027983
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 67.90368482256623,
      "e2e_ms": 61028.4976000039,
      "effective_prefill_tok_s": 3283.189646976845,
      "generation_ms": 677.4301000041305,
      "mean_inter_token_latency_ms": 14.72674130443762,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198144,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 60351.06749999977,
      "ttft_ms": 60351.0682000051,
      "visible_generation_ms": 677.4293999988004
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 62.09271062600077,
      "e2e_ms": 61236.73940000299,
      "effective_prefill_tok_s": 3273.5692891549934,
      "generation_ms": 708.6178000026848,
      "mean_inter_token_latency_ms": 16.10495000006102,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198143,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 60528.121600000304,
      "ttft_ms": 60528.12200000335,
      "visible_generation_ms": 708.617399999639
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260829T233018Z",
  "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": "2026-08-29T23:30:18Z",
  "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": 225431.53170000005
}