{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 5,
  "concurrency": 1,
  "context_distribution": {
    "4096": 5
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-b12x-fixed-mtp5-vision",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T22:18:10Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 66,
    "content_chunks_s": null,
    "decode_tok_s_p50": 64.53947427849076,
    "decode_tok_s_p95": 74.53862027482835,
    "e2e_p50_ms": 1671.9677000000956,
    "e2e_p95_ms": 1799.3405000015628,
    "effective_prefill_tok_s_p50": 2869.743428696412,
    "effective_prefill_tok_s_p95": 3044.9770190465447,
    "generation_p50_ms": 656.2724999967031,
    "generation_p95_ms": 697.2477000017534,
    "mean_inter_token_latency_ms_p50": 15.494393333372297,
    "mean_inter_token_latency_ms_p95": 19.730881249870436,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 207,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 14840,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 6,
    "throughput_tok_s": 24.702088634403882,
    "time_to_first_output_p50_ms": 1034.2387999990024,
    "time_to_first_output_p95_ms": 1167.9523000057088,
    "ttft_p50_ms": 1079.500799998641,
    "ttft_p95_ms": 1167.9533999995328,
    "ttft_samples": 5
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-fixed-mtp5-vision",
  "request_timings": [
    {
      "content_chunks": 14,
      "decode_tok_s": 50.68197346768617,
      "e2e_ms": 1799.3405000015628,
      "effective_prefill_tok_s": 2541.199670556317,
      "generation_ms": 631.3881999958539,
      "mean_inter_token_latency_ms": 19.730881249870436,
      "output_token_source": "usage",
      "output_tokens": 33,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 1167.9523000057088,
      "ttft_ms": 1167.9533999995328,
      "visible_generation_ms": 631.3871000020299
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 60.95029122841647,
      "e2e_ms": 1693.8192000015988,
      "effective_prefill_tok_s": 2860.594130351911,
      "generation_ms": 656.2724999967031,
      "mean_inter_token_latency_ms": 16.406812499917578,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 1037.5467000048957,
      "ttft_ms": 1037.5475000037113,
      "visible_generation_ms": 656.2716999978875
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 64.53947427849076,
      "e2e_ms": 1671.9677000000956,
      "effective_prefill_tok_s": 3044.9770190465447,
      "generation_ms": 697.2477000017534,
      "mean_inter_token_latency_ms": 15.494393333372297,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 974.7199999983422,
      "ttft_ms": 1079.500799998641,
      "visible_generation_ms": 592.4669000014546
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 67.75493863540918,
      "e2e_ms": 1655.8825000029174,
      "effective_prefill_tok_s": 3037.9792442866383,
      "generation_ms": 678.917299999739,
      "mean_inter_token_latency_ms": 14.759071739124762,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 3,
      "time_to_first_output_ms": 976.9652000031783,
      "ttft_ms": 1080.4340000031516,
      "visible_generation_ms": 575.4484999997658
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 74.53862027482835,
      "e2e_ms": 1557.4574999991455,
      "effective_prefill_tok_s": 2869.743428696412,
      "generation_ms": 523.2187000001431,
      "mean_inter_token_latency_ms": 13.415864102567772,
      "output_token_source": "usage",
      "output_tokens": 40,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 1034.2387999990024,
      "ttft_ms": 1034.2392999955337,
      "visible_generation_ms": 523.2182000036119
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260829T221802Z",
  "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": "2026-08-29T22:18:02Z",
  "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": 8379.858199994487
}