{
  "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-b12x-fixed-mtp5-vision",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T22:23:17Z",
  "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": 96,
    "content_chunks_s": null,
    "decode_tok_s_p50": 48.00598927413533,
    "decode_tok_s_p95": 57.3203937114941,
    "e2e_p50_ms": 35351.99039999861,
    "e2e_p95_ms": 36085.12460000202,
    "effective_prefill_tok_s_p50": 3072.2860532609325,
    "effective_prefill_tok_s_p95": 3410.7602894018714,
    "generation_p50_ms": 958.4403999979259,
    "generation_p95_ms": 1025.9345000013127,
    "mean_inter_token_latency_ms_p50": 20.83073414630745,
    "mean_inter_token_latency_ms_p95": 20.937438775536993,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 236,
    "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": 1.3575944168778014,
    "time_to_first_output_p50_ms": 34393.28179999575,
    "time_to_first_output_p95_ms": 35126.68420000409,
    "ttft_p50_ms": 34393.28250000108,
    "ttft_p95_ms": 35126.684700000624,
    "ttft_samples": 5
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-fixed-mtp5-vision",
  "request_timings": [
    {
      "content_chunks": 14,
      "decode_tok_s": 57.3203937114941,
      "e2e_ms": 31781.81659999973,
      "effective_prefill_tok_s": 3410.7602894018714,
      "generation_ms": 802.5066999980481,
      "mean_inter_token_latency_ms": 17.445797826044526,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105663,
      "reasoning_chunks": 5,
      "request_index": 0,
      "time_to_first_output_ms": 30979.30990000168,
      "ttft_ms": 31167.699300000095,
      "visible_generation_ms": 614.1172999996343
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 48.00598927413533,
      "e2e_ms": 35058.442400004424,
      "effective_prefill_tok_s": 3089.194801801231,
      "generation_ms": 854.0600999986054,
      "mean_inter_token_latency_ms": 20.83073414630745,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105664,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 34204.38230000582,
      "ttft_ms": 34204.38529999956,
      "visible_generation_ms": 854.0571000048658
    },
    {
      "content_chunks": 21,
      "decode_tok_s": 51.11042083053555,
      "e2e_ms": 35351.99039999861,
      "effective_prefill_tok_s": 3072.2860532609325,
      "generation_ms": 958.7086000028648,
      "mean_inter_token_latency_ms": 19.565481632711528,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105666,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 34393.28179999575,
      "ttft_ms": 34393.28250000108,
      "visible_generation_ms": 958.7078999975347
    },
    {
      "content_chunks": 22,
      "decode_tok_s": 47.76133369131977,
      "e2e_ms": 35549.64469999686,
      "effective_prefill_tok_s": 3060.7370814975043,
      "generation_ms": 1025.9345000013127,
      "mean_inter_token_latency_ms": 20.937438775536993,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105668,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 34523.71019999555,
      "ttft_ms": 34523.710599998594,
      "visible_generation_ms": 1025.934099998267
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 47.994637955682535,
      "e2e_ms": 36085.12460000202,
      "effective_prefill_tok_s": 3008.1689293061054,
      "generation_ms": 958.4403999979259,
      "mean_inter_token_latency_ms": 20.835660869520126,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 131072,
      "prompt_tokens": 105667,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 35126.68420000409,
      "ttft_ms": 35126.684700000624,
      "visible_generation_ms": 958.4399000013946
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260829T222023Z",
  "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": "2026-08-29T22:20:23Z",
  "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": 173836.89639999648
}