{
  "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-k3-b12x-nospec-xgrammar-fixed",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-31T08:08:55Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 524288,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 151,
    "content_chunks_s": null,
    "decode_tok_s_p50": 42.60788167760298,
    "decode_tok_s_p95": 43.48941685082938,
    "e2e_p50_ms": 1774.503900000127,
    "e2e_p95_ms": 7435.887600004207,
    "effective_prefill_tok_s_p50": 3192.806117234987,
    "effective_prefill_tok_s_p95": 3265.852976255335,
    "generation_p50_ms": 844.9140999873634,
    "generation_p95_ms": 873.519499989925,
    "mean_inter_token_latency_ms_p50": 23.469836110760095,
    "mean_inter_token_latency_ms_p95": 24.264430555275695,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 175,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 14840,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 14,
    "throughput_tok_s": 12.251582708456569,
    "time_to_first_output_p50_ms": 929.5898000127636,
    "time_to_first_output_p95_ms": 6565.288400015561,
    "ttft_p50_ms": 932.5836999923922,
    "ttft_p95_ms": 6754.639300022973,
    "ttft_samples": 5
  },
  "model": "glm53-flash-exl3-k3-nospec-fp8-tp2-524k-vision-xgfix",
  "request_timings": [
    {
      "content_chunks": 28,
      "decode_tok_s": 41.35083055494363,
      "e2e_ms": 7435.887600004207,
      "effective_prefill_tok_s": 452.07458060684206,
      "generation_ms": 870.5991999886464,
      "mean_inter_token_latency_ms": 24.183311110795735,
      "output_token_source": "usage",
      "output_tokens": 37,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 7,
      "request_index": 0,
      "time_to_first_output_ms": 6565.288400015561,
      "ttft_ms": 6754.639300022973,
      "visible_generation_ms": 681.2482999812346
    },
    {
      "content_chunks": 32,
      "decode_tok_s": 43.48941685082938,
      "e2e_ms": 1685.8491000020877,
      "effective_prefill_tok_s": 3201.574725994455,
      "generation_ms": 758.8052999926731,
      "mean_inter_token_latency_ms": 22.994099999777973,
      "output_token_source": "usage",
      "output_tokens": 34,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 927.0438000094146,
      "ttft_ms": 927.0465000008699,
      "visible_generation_ms": 758.8026000012178
    },
    {
      "content_chunks": 28,
      "decode_tok_s": 41.21258884365514,
      "e2e_ms": 1782.31730000698,
      "effective_prefill_tok_s": 3265.852976255335,
      "generation_ms": 873.519499989925,
      "mean_inter_token_latency_ms": 24.264430555275695,
      "output_token_source": "usage",
      "output_tokens": 37,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 7,
      "request_index": 2,
      "time_to_first_output_ms": 908.7978000170551,
      "ttft_ms": 1101.5369000087958,
      "visible_generation_ms": 680.7803999981843
    },
    {
      "content_chunks": 28,
      "decode_tok_s": 43.199002310378305,
      "e2e_ms": 1603.8946000044234,
      "effective_prefill_tok_s": 3182.559232323417,
      "generation_ms": 671.3118000188842,
      "mean_inter_token_latency_ms": 23.14868275927187,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 932.5827999855392,
      "ttft_ms": 932.5836999923922,
      "visible_generation_ms": 671.3109000120312
    },
    {
      "content_chunks": 35,
      "decode_tok_s": 42.60788167760298,
      "e2e_ms": 1774.503900000127,
      "effective_prefill_tok_s": 3192.806117234987,
      "generation_ms": 844.9140999873634,
      "mean_inter_token_latency_ms": 23.469836110760095,
      "output_token_source": "usage",
      "output_tokens": 37,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 929.5898000127636,
      "ttft_ms": 929.5904000173323,
      "visible_generation_ms": 844.9134999827947
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260831T080841Z",
  "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": "2026-08-31T08:08:41Z",
  "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": 14283.868800004711
}