{
  "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-k3-b12x-dflash2-k5",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T23:45:00Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 1048576,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 95,
    "content_chunks_s": null,
    "decode_tok_s_p50": 25.89591352123513,
    "decode_tok_s_p95": 69.69046696501046,
    "e2e_p50_ms": 2385.3384999965783,
    "e2e_p95_ms": 3717.0761000015773,
    "effective_prefill_tok_s_p50": 2759.480157826221,
    "effective_prefill_tok_s_p95": 2911.6079767039296,
    "generation_p50_ms": 1351.5646000014385,
    "generation_p95_ms": 1478.7474000040675,
    "mean_inter_token_latency_ms_p50": 36.61132000015641,
    "mean_inter_token_latency_ms_p95": 42.91280322579569,
    "output_token_sources": {
      "usage": 8
    },
    "output_tokens": 287,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 8,
    "prompt_tokens": 23744,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 9,
    "throughput_tok_s": 28.788435771605098,
    "time_to_first_output_p50_ms": 1063.2188999952632,
    "time_to_first_output_p95_ms": 2238.32869999751,
    "ttft_p50_ms": 1075.5656999972416,
    "ttft_p95_ms": 2238.3305999974255,
    "ttft_samples": 8
  },
  "model": "glm53-flash-exl3-k3-dflash2-k5-fp8-tp2-1m-vision",
  "request_timings": [
    {
      "content_chunks": 10,
      "decode_tok_s": 27.71584461586692,
      "e2e_ms": 2613.5753999988083,
      "effective_prefill_tok_s": 2388.6968282648627,
      "generation_ms": 1371.0569000031683,
      "mean_inter_token_latency_ms": 36.08044473692548,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 0,
      "time_to_first_output_ms": 1242.51849999564,
      "ttft_ms": 2198.7312999990536,
      "visible_generation_ms": 414.8440999997547
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 28.402416802142458,
      "e2e_ms": 3717.0761000015773,
      "effective_prefill_tok_s": 1325.9893419600535,
      "generation_ms": 1478.7474000040675,
      "mean_inter_token_latency_ms": 35.208271428668276,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 2238.32869999751,
      "ttft_ms": 2238.3305999974255,
      "visible_generation_ms": 1478.7455000041518
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 25.89591352123513,
      "e2e_ms": 2370.9326000025612,
      "effective_prefill_tok_s": 2911.6079767039296,
      "generation_ms": 1351.5646000014385,
      "mean_inter_token_latency_ms": 38.616131428612526,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 1019.3680000011227,
      "ttft_ms": 1103.7901000017882,
      "visible_generation_ms": 1267.142500000773
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 23.303068660843884,
      "e2e_ms": 2385.3384999965783,
      "effective_prefill_tok_s": 2813.1592157206765,
      "generation_ms": 1330.2968999996665,
      "mean_inter_token_latency_ms": 42.91280322579569,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 1055.0415999969118,
      "ttft_ms": 1055.042299994966,
      "visible_generation_ms": 1330.2962000016123
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 23.546583561957004,
      "e2e_ms": 2477.0420999993803,
      "effective_prefill_tok_s": 2759.480157826221,
      "generation_ms": 1401.4772000009543,
      "mean_inter_token_latency_ms": 42.46900606063498,
      "output_token_source": "usage",
      "output_tokens": 34,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 1075.564899998426,
      "ttft_ms": 1075.5656999972416,
      "visible_generation_ms": 1401.4764000021387
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 23.397207370293792,
      "e2e_ms": 2046.2421999982325,
      "effective_prefill_tok_s": 2791.5229874235897,
      "generation_ms": 983.0233000029693,
      "mean_inter_token_latency_ms": 42.740143478389975,
      "output_token_source": "usage",
      "output_tokens": 24,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 1063.2188999952632,
      "ttft_ms": 1063.2193999990704,
      "visible_generation_ms": 983.0227999991621
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 27.31395644832603,
      "e2e_ms": 2506.280500005232,
      "effective_prefill_tok_s": 2848.839592192565,
      "generation_ms": 1464.4528000062564,
      "mean_inter_token_latency_ms": 36.61132000015641,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 1041.8276999989757,
      "ttft_ms": 1041.828200002783,
      "visible_generation_ms": 1464.4523000024492
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 69.69046696501046,
      "e2e_ms": 1818.6283000031835,
      "effective_prefill_tok_s": 2304.8682109318365,
      "generation_ms": 530.9191000051214,
      "mean_inter_token_latency_ms": 14.349164865003281,
      "output_token_source": "usage",
      "output_tokens": 38,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 1287.709199998062,
      "ttft_ms": 1378.2214000020758,
      "visible_generation_ms": 440.4069000011077
    }
  ],
  "requests": 8,
  "run_id": "benchmark-20260829T234450Z",
  "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": "2026-08-29T23:44:50Z",
  "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": 9969.280800003617
}