{
  "base_url": "http://127.0.0.1:39081/v1",
  "completed": 5,
  "concurrency": 1,
  "context_distribution": {
    "4096": 5
  },
  "context_seed": 38,
  "context_tokens": 4096,
  "engine": "sglang-d91c3682b-sm120-qsa-pr36556-mtp3",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-27T03:42:41Z",
  "gpu": "2x-rtx-pro-6000-blackwell-max-q-tp2-pcie-wsl2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 512,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 56,
    "content_chunks_s": null,
    "decode_tok_s_p50": 155.85024105965064,
    "decode_tok_s_p95": 160.26660083630853,
    "e2e_p50_ms": 339.8921999614686,
    "e2e_p95_ms": 861.6299999412149,
    "effective_prefill_tok_s_p50": 25627.148738185137,
    "effective_prefill_tok_s_p95": 26467.449001030778,
    "generation_p50_ms": 198.908899910748,
    "generation_p95_ms": 219.09269993193448,
    "mean_inter_token_latency_ms_p50": 6.416416126153162,
    "mean_inter_token_latency_ms_p95": 7.303089997731149,
    "output_token_sources": {
      "usage": 5
    },
    "output_tokens": 159,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 5,
    "prompt_tokens": 18065,
    "prompt_tokens_p50": 3613.0,
    "prompt_tokens_p95": 3613.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 71.44426605761159,
    "time_to_first_output_p50_ms": 140.98330005072057,
    "time_to_first_output_p95_ms": 642.5373000092804,
    "ttft_p50_ms": 140.98370005376637,
    "ttft_p95_ms": 642.5379998981953,
    "ttft_samples": 5
  },
  "model": "qwen38-flash-next-radixark-nvfp4-sglang-qsa-fast-tp2-262k-mtp3",
  "request_timings": [
    {
      "content_chunks": 12,
      "decode_tok_s": 136.92834133369163,
      "e2e_ms": 861.6299999412149,
      "effective_prefill_tok_s": 5623.019861956365,
      "generation_ms": 219.09269993193448,
      "mean_inter_token_latency_ms": 7.303089997731149,
      "output_token_source": "usage",
      "output_tokens": 31,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 642.5373000092804,
      "ttft_ms": 642.5379998981953,
      "visible_generation_ms": 219.09200004301965
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 158.74828516059105,
      "e2e_ms": 337.89269998669624,
      "effective_prefill_tok_s": 25333.941026697445,
      "generation_ms": 195.27769996784627,
      "mean_inter_token_latency_ms": 6.299280644124074,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 142.61500001884997,
      "ttft_ms": 142.61540002189577,
      "visible_generation_ms": 195.27729996480048
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 144.17826763183587,
      "e2e_ms": 354.7851999755949,
      "effective_prefill_tok_s": 25848.944286061476,
      "generation_ms": 215.0115999393165,
      "mean_inter_token_latency_ms": 6.935858062558597,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 139.77360003627837,
      "ttft_ms": 139.77410015650094,
      "visible_generation_ms": 215.01109981909394
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 155.85024105965064,
      "e2e_ms": 339.8921999614686,
      "effective_prefill_tok_s": 25627.148738185137,
      "generation_ms": 198.908899910748,
      "mean_inter_token_latency_ms": 6.416416126153162,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 140.98330005072057,
      "ttft_ms": 140.98370005376637,
      "visible_generation_ms": 198.9084999077022
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 160.26660083630853,
      "e2e_ms": 329.935000045225,
      "effective_prefill_tok_s": 26467.449001030778,
      "generation_ms": 193.4277000837028,
      "mean_inter_token_latency_ms": 6.239603228506542,
      "output_token_source": "usage",
      "output_tokens": 32,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 136.50729996152222,
      "ttft_ms": 136.50769996456802,
      "visible_generation_ms": 193.427300080657
    }
  ],
  "requests": 5,
  "run_id": "benchmark-20260827T034239Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": null,
    "shared_prefix_burst": false,
    "thinking_mode": "disabled"
  },
  "source_recipe": null,
  "started_at": "2026-08-27T03:42:39Z",
  "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": 2225.5110000260174
}