{
  "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-b12x-fixed-mtp5-vision",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T22:28:44Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 100,
    "content_chunks_s": null,
    "decode_tok_s_p50": 45.714369306151546,
    "decode_tok_s_p95": 49.68531803813854,
    "e2e_p50_ms": 65324.84850000037,
    "e2e_p95_ms": 65613.96420000528,
    "effective_prefill_tok_s_p50": 3080.316435649605,
    "effective_prefill_tok_s_p95": 3090.443431442682,
    "generation_p50_ms": 1006.3335000013467,
    "generation_p95_ms": 1100.1424999994924,
    "mean_inter_token_latency_ms_p50": 21.874960000059218,
    "mean_inter_token_latency_ms_p95": 22.00284999998985,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 221,
    "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": 0.6775335516799121,
    "time_to_first_output_p50_ms": 64324.56020000245,
    "time_to_first_output_p95_ms": 64519.961900005,
    "ttft_p50_ms": 64324.560900000506,
    "ttft_p95_ms": 64519.962600003055,
    "ttft_samples": 5
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-fixed-mtp5-vision",
  "request_timings": [
    {
      "content_chunks": 10,
      "decode_tok_s": 45.714369306151546,
      "e2e_ms": 64550.95230000006,
      "effective_prefill_tok_s": 3090.443431442682,
      "generation_ms": 437.49920000118436,
      "mean_inter_token_latency_ms": 21.874960000059218,
      "output_token_source": "usage",
      "output_tokens": 21,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198139,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 64113.45309999888,
      "ttft_ms": 64113.45419999998,
      "visible_generation_ms": 437.4981000000844
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 45.986742022370635,
      "e2e_ms": 65324.84850000037,
      "effective_prefill_tok_s": 3080.316435649605,
      "generation_ms": 1000.2882999979192,
      "mean_inter_token_latency_ms": 21.74539782604172,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198140,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 64324.56020000245,
      "ttft_ms": 64324.560900000506,
      "visible_generation_ms": 1000.287599999865
    },
    {
      "content_chunks": 24,
      "decode_tok_s": 45.448657787534856,
      "e2e_ms": 65217.32019999763,
      "effective_prefill_tok_s": 3090.310695319419,
      "generation_ms": 1100.1424999994924,
      "mean_inter_token_latency_ms": 22.00284999998985,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198142,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 64117.17769999814,
      "ttft_ms": 64117.178200001945,
      "visible_generation_ms": 1100.1419999956852
    },
    {
      "content_chunks": 23,
      "decode_tok_s": 45.703743035994556,
      "e2e_ms": 65613.96420000528,
      "effective_prefill_tok_s": 3071.049550635036,
      "generation_ms": 1094.0023000002839,
      "mean_inter_token_latency_ms": 21.880046000005677,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198144,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 64519.961900005,
      "ttft_ms": 64519.962600003055,
      "visible_generation_ms": 1094.0016000022297
    },
    {
      "content_chunks": 22,
      "decode_tok_s": 49.68531803813854,
      "e2e_ms": 65457.49619999697,
      "effective_prefill_tok_s": 3074.31226527762,
      "generation_ms": 1006.3335000013467,
      "mean_inter_token_latency_ms": 20.126670000026934,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 240000,
      "prompt_tokens": 198143,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 64451.16269999562,
      "ttft_ms": 64451.16319999943,
      "visible_generation_ms": 1006.3329999975394
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260829T222318Z",
  "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": "2026-08-29T22:23:18Z",
  "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": 326183.10850000125
}