{
 "schema": 1,
 "suite": "latency-triage",
 "date": "2026-10-06T08:07:49Z",
 "ollaya": "0.10.0",
 "machine": "rtx-4090",
 "machine_detail": {
  "id": "rtx-4090",
  "cpu": "13th Gen Intel(R) Core(TM) i9-13900K",
  "threads": 24,
  "gpus": [
   {
    "name": "NVIDIA GeForce RTX 4090",
    "driver": "616.92",
    "memory": "24564 MiB"
   }
  ],
  "os": "Linux 6.18.33.2-microsoft-standard-WSL2"
 },
 "device": "cuda",
 "protocol": {
  "preset": "triage",
  "questions": 5,
  "states": 30,
  "warm": 5,
  "n": 20,
  "client": "one request at a time over HTTP to /api/decide, timed by the client; each request a different short message, in turn"
 },
 "note": "snap, added with Ollaya 0.10.0; the same protocol and machine as the 2026-10-01 RTX 4090 sweep.",
 "results": [
  {
   "model": "snap:2b",
   "questions": "triage",
   "load_ms": 635.9,
   "first_request_ms": 704.7,
   "answered_by": "snap:2b",
   "p50_ms": 67.3,
   "p90_ms": 68.2,
   "min_ms": 66.6,
   "eval_p50_ms": 66.7,
   "input_tokens": 606,
   "device": "cuda:0",
   "precision": "Q8_0"
  }
 ]
}