{
  "source": "Frozen authenticated exact fast r5 / pro r3 launch mix c2",
  "measured": "2026-09-27",
  "source_receipt": "launch-r5-mix-c2.json",
  "source_receipt_sha256": "ee0c79fd2cd4bf9f5a1f78774356535519a1a6519dd47b7619eb1c2d6587b684",
  "generator_sha256": "efdd3d3f61dcc97cd685e9c5e0ad0151d2b11d20febabe14b206152fc191624e",
  "environment": "Local development GPU test; one NVIDIA RTX PRO 6000 96GB; exact s1-fast r5 and s1-pro r3 images; internal engine authentication enabled",
  "workload": {
    "concurrency": 2,
    "submitted_requests": 1000,
    "mix": "700 s1-fast text requests; 300 s1-pro requests, including 60 images; s1-pro text:image ratio 4:1",
    "inputs": "Public benchmark inputs only"
  },
  "results": {
    "s1-fast": {
      "submitted": 700,
      "completed": 700,
      "p50_latency_ms": 36.39949578791857,
      "p95_latency_ms": 243.54618974030018,
      "errors": 0
    },
    "s1-pro": {
      "submitted": 300,
      "completed": 300,
      "p50_latency_ms": 208.11019744724035,
      "p95_latency_ms": 644.4333009421825,
      "images_completed": 60,
      "errors": 0
    }
  },
  "limitations": "Local model-server measurement only. It is not a production, Verda, or end-to-end service measurement, service-level promise, or latency guarantee.",
  "measurement_scope": "Model-server latency; excludes API gateway and network time.",
  "source_hashes": {
    "JevBench public text inputs": "00922b026c13ad37a5a49e14c31228016d3e434537b12d7dfa0ba50355ee85da",
    "ImageJevBench public image inputs": "3eb83cede6feeb6585ceb3925653c830112fd69e4645761c3995ea34340ce390"
  }
}