{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 8,
  "concurrency": 2,
  "context_distribution": {
    "4096": 8
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-b12x-fixed-mtp5-vision",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T22:29:07Z",
  "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": 113,
    "content_chunks_s": null,
    "decode_tok_s_p50": 33.29814336214366,
    "decode_tok_s_p95": 55.65684217370226,
    "e2e_p50_ms": 2712.7672000060556,
    "e2e_p95_ms": 3434.7250999999233,
    "effective_prefill_tok_s_p50": 1755.4903582044938,
    "effective_prefill_tok_s_p95": 2779.44157198254,
    "generation_p50_ms": 1006.1656000034418,
    "generation_p95_ms": 2016.1091999980272,
    "mean_inter_token_latency_ms_p50": 29.205086206965117,
    "mean_inter_token_latency_ms_p95": 39.88742413815183,
    "output_token_sources": {
      "usage": 8
    },
    "output_tokens": 353,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 8,
    "prompt_tokens": 23744,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 15,
    "throughput_tok_s": 30.791493508861684,
    "time_to_first_output_p50_ms": 1372.4307999946177,
    "time_to_first_output_p95_ms": 2277.98979999352,
    "ttft_p50_ms": 1743.2322000022396,
    "ttft_p95_ms": 2277.9904999988503,
    "ttft_samples": 8
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-fixed-mtp5-vision",
  "request_timings": [
    {
      "content_chunks": 11,
      "decode_tok_s": 25.07055849323477,
      "e2e_ms": 3434.7250999999233,
      "effective_prefill_tok_s": 1302.9031122125493,
      "generation_ms": 1156.7353000064031,
      "mean_inter_token_latency_ms": 39.88742413815183,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 2277.98979999352,
      "ttft_ms": 2277.9904999988503,
      "visible_generation_ms": 1156.734600001073
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 34.240611135792854,
      "e2e_ms": 2712.7672000060556,
      "effective_prefill_tok_s": 1590.7217615901097,
      "generation_ms": 846.9475000019884,
      "mean_inter_token_latency_ms": 29.205086206965117,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 1865.8197000040673,
      "ttft_ms": 1865.821200000937,
      "visible_generation_ms": 846.9460000051185
    },
    {
      "content_chunks": 20,
      "decode_tok_s": 28.272278108772966,
      "e2e_ms": 3083.9495999971405,
      "effective_prefill_tok_s": 2779.44157198254,
      "generation_ms": 2016.1091999980272,
      "mean_inter_token_latency_ms": 35.37033684207065,
      "output_token_source": "usage",
      "output_tokens": 58,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 1067.8403999991133,
      "ttft_ms": 1743.2322000022396,
      "visible_generation_ms": 1340.717399994901
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 50.56951954811026,
      "e2e_ms": 2259.4476000012946,
      "effective_prefill_tok_s": 2167.0821822783455,
      "generation_ms": 889.8640999977943,
      "mean_inter_token_latency_ms": 19.774757777728762,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 3,
      "time_to_first_output_ms": 1369.5835000035004,
      "ttft_ms": 1488.8986000005389,
      "visible_generation_ms": 770.5490000007558
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 40.2239670484028,
      "e2e_ms": 2660.2668999985326,
      "effective_prefill_tok_s": 1755.4903582044938,
      "generation_ms": 969.5712000029744,
      "mean_inter_token_latency_ms": 24.860800000076267,
      "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": 1690.6956999955582,
      "ttft_ms": 1690.6961999993655,
      "visible_generation_ms": 969.5706999991671
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 33.29814336214366,
      "e2e_ms": 3287.2901000009733,
      "effective_prefill_tok_s": 1533.1660281387549,
      "generation_ms": 1351.4266999991378,
      "mean_inter_token_latency_ms": 30.031704444425284,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 5,
      "time_to_first_output_ms": 1935.8634000018355,
      "ttft_ms": 2060.01240000478,
      "visible_generation_ms": 1227.2776999961934
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 26.397626747773757,
      "e2e_ms": 2779.7082000033697,
      "effective_prefill_tok_s": 2760.9058338452933,
      "generation_ms": 1704.698699999426,
      "mean_inter_token_latency_ms": 37.88219333332058,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 6,
      "time_to_first_output_ms": 1075.0095000039437,
      "ttft_ms": 1755.786800000351,
      "visible_generation_ms": 1023.9214000030188
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 55.65684217370226,
      "e2e_ms": 2378.5963999980595,
      "effective_prefill_tok_s": 2162.586266653036,
      "generation_ms": 1006.1656000034418,
      "mean_inter_token_latency_ms": 17.96724285720432,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 1372.4307999946177,
      "ttft_ms": 1492.2391000000061,
      "visible_generation_ms": 886.3572999980533
    }
  ],
  "requests": 8,
  "run_id": "benchmark-20260829T222855Z",
  "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": "2026-08-29T22:28:55Z",
  "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": 11464.205199998105
}