{
  "base_url": "http://127.0.0.1:39052/v1",
  "completed": 4,
  "concurrency": 1,
  "context_distribution": {
    "60000": 4
  },
  "context_seed": 0,
  "context_tokens": 60000,
  "engine": "nvidia-vllm-0.22.1",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-01T12:53:11Z",
  "gpu": "dual-rtx-pro-6000-blackwell-max-q-tp2-pcie",
  "manifest": null,
  "max_context_tokens": 65536,
  "max_tokens": 128,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 236,
    "content_chunks_s": null,
    "decode_tok_s_p50": 59.96134427523613,
    "decode_tok_s_p95": 60.356575149460085,
    "e2e_p50_ms": 6612.638500000685,
    "e2e_p95_ms": 6632.338299999901,
    "effective_prefill_tok_s_p50": 9645.846007097885,
    "effective_prefill_tok_s_p95": 9652.390727817594,
    "generation_p50_ms": 1033.9995000013005,
    "generation_p95_ms": 1052.7344000001904,
    "mean_inter_token_latency_ms_p50": 16.65363541667375,
    "mean_inter_token_latency_ms_p95": 16.710069841272862,
    "output_token_sources": {
      "usage": 4
    },
    "output_tokens": 240,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 4,
    "prompt_tokens": 215285,
    "prompt_tokens_p50": 53820.0,
    "prompt_tokens_p95": 53825.0,
    "throughput_tok_s": 9.131511005622068,
    "ttft_p50_ms": 5578.638999999384,
    "ttft_p95_ms": 5613.548900000751,
    "ttft_samples": 4
  },
  "model": "nemotron3-super-120b-a12b-nvfp4-tp2",
  "request_timings": [
    {
      "decode_tok_s": 60.046949208386785,
      "e2e_ms": 6412.923400001091,
      "effective_prefill_tok_s": 9587.161519158193,
      "generation_ms": 799.37450000034,
      "mean_inter_token_latency_ms": 16.65363541667375,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 60000,
      "prompt_tokens": 53818,
      "request_index": 0,
      "ttft_ms": 5613.548900000751
    },
    {
      "decode_tok_s": 59.84415442298514,
      "e2e_ms": 6632.338299999901,
      "effective_prefill_tok_s": 9645.846007097885,
      "generation_ms": 1052.7344000001904,
      "mean_inter_token_latency_ms": 16.710069841272862,
      "output_token_source": "usage",
      "output_tokens": 64,
      "planned_context_tokens": 60000,
      "prompt_tokens": 53820,
      "request_index": 1,
      "ttft_ms": 5579.603899999711
    },
    {
      "decode_tok_s": 60.356575149460085,
      "e2e_ms": 6619.824699999299,
      "effective_prefill_tok_s": 9652.390727817594,
      "generation_ms": 1043.796800000564,
      "mean_inter_token_latency_ms": 16.568203174612126,
      "output_token_source": "usage",
      "output_tokens": 64,
      "planned_context_tokens": 60000,
      "prompt_tokens": 53822,
      "request_index": 2,
      "ttft_ms": 5576.027899998735
    },
    {
      "decode_tok_s": 59.96134427523613,
      "e2e_ms": 6612.638500000685,
      "effective_prefill_tok_s": 9648.410660737492,
      "generation_ms": 1033.9995000013005,
      "mean_inter_token_latency_ms": 16.677411290343557,
      "output_token_source": "usage",
      "output_tokens": 63,
      "planned_context_tokens": 60000,
      "prompt_tokens": 53825,
      "request_index": 3,
      "ttft_ms": 5578.638999999384
    }
  ],
  "requests": 4,
  "run_id": "benchmark-20260801T125245Z",
  "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": "2026-08-01T12:52:45Z",
  "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": 26282.616300000882
}