{
  "base_url": "http://127.0.0.1:30010/v1",
  "completed": 2,
  "concurrency": 1,
  "context_distribution": {
    "240000": 2
  },
  "context_seed": 0,
  "context_tokens": 240000,
  "engine": "vllm-f25953cc",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-07-29T10:25:14Z",
  "gpu": "rtx-pro-6000",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 64,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 120,
    "content_chunks_s": null,
    "decode_tok_s_p50": 155.84041644630634,
    "decode_tok_s_p95": 157.19006049793688,
    "e2e_p50_ms": 33375.483600000734,
    "e2e_p95_ms": 33846.54739999678,
    "effective_prefill_tok_s_p50": 6920.220353226669,
    "effective_prefill_tok_s_p95": 7018.290865217369,
    "generation_p50_ms": 400.7887000007031,
    "generation_p95_ms": 404.2596999970556,
    "mean_inter_token_latency_ms_p50": 6.361725396836557,
    "mean_inter_token_latency_ms_p95": 6.416820634873898,
    "output_token_sources": {
      "usage": 2
    },
    "output_tokens": 128,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 2,
    "prompt_tokens": 462854,
    "prompt_tokens_p50": 231426.0,
    "prompt_tokens_p95": 231428.0,
    "throughput_tok_s": 1.903913592261946,
    "ttft_p50_ms": 32974.69490000003,
    "ttft_p95_ms": 33442.287699999724,
    "ttft_samples": 2
  },
  "model": "agents-a1-fp8-mm-262k",
  "request_timings": [
    {
      "decode_tok_s": 157.19006049793688,
      "e2e_ms": 33375.483600000734,
      "effective_prefill_tok_s": 7018.290865217369,
      "generation_ms": 400.7887000007031,
      "mean_inter_token_latency_ms": 6.361725396836557,
      "output_token_source": "usage",
      "output_tokens": 64,
      "planned_context_tokens": 240000,
      "prompt_tokens": 231426,
      "request_index": 0,
      "ttft_ms": 32974.69490000003
    },
    {
      "decode_tok_s": 155.84041644630634,
      "e2e_ms": 33846.54739999678,
      "effective_prefill_tok_s": 6920.220353226669,
      "generation_ms": 404.2596999970556,
      "mean_inter_token_latency_ms": 6.416820634873898,
      "output_token_source": "usage",
      "output_tokens": 64,
      "planned_context_tokens": 240000,
      "prompt_tokens": 231428,
      "request_index": 1,
      "ttft_ms": 33442.287699999724
    }
  ],
  "requests": 2,
  "run_id": "benchmark-20260729T102407Z",
  "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": "2026-07-29T10:24:07Z",
  "tier": null,
  "timing_methodology": {
    "clock": "client time.perf_counter",
    "decode": "usage completion tokens after the first token divided by client-observed generation time",
    "effective_prefill": "usage.prompt_tokens divided by client-observed TTFT; includes queueing, scheduling, prefill, and first-token work",
    "generation": "client-observed E2E minus TTFT",
    "mean_inter_token_latency": "client-observed generation time divided by completion tokens after the first token; not a raw per-token timestamp trace",
    "ttft": "request start through the first non-empty streamed content delta"
  },
  "wall_clock_ms": 67229.94180000023
}