{
  "base_url": "http://127.0.0.1:39069/v1",
  "completed": 4,
  "concurrency": 1,
  "context_distribution": {
    "150000": 6
  },
  "context_seed": 0,
  "context_tokens": 150000,
  "engine": "vllm-r16-b12x-wsl2-offload-mmap-unpinned",
  "failed": 2,
  "failures": [
    {
      "error_message": "stream completed without visible content",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "request_index": 1
    },
    {
      "error_message": "stream completed without visible content",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "request_index": 4
    }
  ],
  "finished_at": "2026-08-02T06:18:37Z",
  "gpu": "2x RTX PRO 6000 Blackwell Max-Q TP2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 1024,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 31,
    "content_chunks_s": null,
    "decode_tok_s_p50": 113.9047213113628,
    "decode_tok_s_p95": 135.73084544482919,
    "e2e_p50_ms": 19873.598100006348,
    "e2e_p95_ms": 22218.934000004083,
    "effective_prefill_tok_s_p50": 6319.898515127279,
    "effective_prefill_tok_s_p95": 6958.668581637918,
    "generation_p50_ms": 1887.0751000067685,
    "generation_p95_ms": 4328.178800002206,
    "mean_inter_token_latency_ms_p50": 8.500338288318776,
    "mean_inter_token_latency_ms_p95": 9.03677867641176,
    "output_token_sources": {
      "usage": 4
    },
    "output_tokens": 1231,
    "prefix_cache_hit_avg": 0.03715889297656392,
    "prompt_token_samples": 4,
    "prompt_tokens": 454696,
    "prompt_tokens_p50": 113674.0,
    "prompt_tokens_p95": 113675.0,
    "reasoning_chunks": 347,
    "throughput_tok_s": 9.449013159151491,
    "time_to_first_output_p50_ms": 17890.755200001877,
    "time_to_first_output_p95_ms": 18980.866900004912,
    "ttft_p50_ms": 19542.319600004703,
    "ttft_p95_ms": 22046.91129999992,
    "ttft_samples": 4
  },
  "model": "deepseek-v4-flash-0731-r16-b12x-dspark5-offload16-wsl2-mmap-unpinned-tp2-256k",
  "request_timings": [
    {
      "content_chunks": 7,
      "decode_tok_s": 110.65890134171912,
      "e2e_ms": 20209.86879999691,
      "effective_prefill_tok_s": 5988.87293182434,
      "generation_ms": 1229.0018999919994,
      "mean_inter_token_latency_ms": 9.03677867641176,
      "output_token_source": "usage",
      "output_tokens": 137,
      "planned_context_tokens": 150000,
      "prompt_tokens": 113674,
      "reasoning_chunks": 38,
      "request_index": 0,
      "time_to_first_output_ms": 18980.866900004912,
      "ttft_ms": 20035.06200000993,
      "visible_generation_ms": 174.8067999869818
    },
    {
      "content_chunks": 7,
      "decode_tok_s": 113.9047213113628,
      "e2e_ms": 22218.934000004083,
      "effective_prefill_tok_s": 6353.784327672656,
      "generation_ms": 4328.178800002206,
      "mean_inter_token_latency_ms": 8.779267342803664,
      "output_token_source": "usage",
      "output_tokens": 494,
      "planned_context_tokens": 150000,
      "prompt_tokens": 113674,
      "reasoning_chunks": 148,
      "request_index": 2,
      "time_to_first_output_ms": 17890.755200001877,
      "ttft_ms": 22046.91129999992,
      "visible_generation_ms": 172.0227000041632
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 117.64237681860342,
      "e2e_ms": 19873.598100006348,
      "effective_prefill_tok_s": 6319.898515127279,
      "generation_ms": 1887.0751000067685,
      "mean_inter_token_latency_ms": 8.500338288318776,
      "output_token_source": "usage",
      "output_tokens": 223,
      "planned_context_tokens": 150000,
      "prompt_tokens": 113673,
      "reasoning_chunks": 56,
      "request_index": 3,
      "time_to_first_output_ms": 17986.52299999958,
      "ttft_ms": 19542.319600004703,
      "visible_generation_ms": 331.2785000016447
    },
    {
      "content_chunks": 5,
      "decode_tok_s": 135.73084544482919,
      "e2e_ms": 19105.92820000602,
      "effective_prefill_tok_s": 6958.668581637918,
      "generation_ms": 2770.188299997244,
      "mean_inter_token_latency_ms": 7.367522074460755,
      "output_token_source": "usage",
      "output_tokens": 377,
      "planned_context_tokens": 150000,
      "prompt_tokens": 113675,
      "reasoning_chunks": 105,
      "request_index": 5,
      "time_to_first_output_ms": 16335.739900008775,
      "ttft_ms": 19000.36840001121,
      "visible_generation_ms": 105.55979999480769
    }
  ],
  "requests": 6,
  "run_id": "benchmark-20260802T061627Z",
  "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": "2026-08-02T06:16:27Z",
  "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": 130278.15489999193
}