{
 "schema": 1,
 "suite": "latency-triage",
 "date": "2026-10-19T11:14:03Z",
 "ollaya": "0.14.0",
 "machine": "rtx-5090",
 "machine_detail": {
  "id": "rtx-5090",
  "cpu": "AMD Ryzen Threadripper 3960X 24-Core Processor",
  "threads": 48,
  "gpus": [
   {
    "name": "NVIDIA GeForce RTX 5090",
    "driver": "596.21",
    "memory": "32607 MiB"
   }
  ],
  "os": "Linux 6.6.87.2-microsoft-standard-WSL2"
 },
 "device": "cpu",
 "protocol": {
  "preset": "triage",
  "questions": 5,
  "states": 30,
  "warm": 2,
  "n": 10,
  "client": "one request at a time over HTTP to /api/decide, timed by the client; each request a different short message, in turn",
  "memory": "vram_mb: peak rise of the GPU's used memory over its idle level before the load (nvidia-smi, 100 ms); ram_peak_mb and ram_mb: the runner's VmHWM and VmRSS"
 },
 "note": "Ollaya 0.14.0. Release candidate (feat/fp16-decoders), measured before the release; one model loaded at a time.",
 "results": [
  {
   "model": "nimble:9b-v3",
   "questions": "triage",
   "load_ms": 19115.2,
   "first_request_ms": 71569.8,
   "answered_by": "nimble:9b-v3",
   "p50_ms": 50928.2,
   "p90_ms": 51052.6,
   "min_ms": 50614.2,
   "eval_p50_ms": 50927.1,
   "input_tokens": 2798,
   "device": "cpu",
   "precision": "F32",
   "digest": "01cb73cdc839e8eb25a6209976a2cd1e83048b6b893c809d7144e71b34ff2b31",
   "size_mb": 19105,
   "ram_peak_mb": 19256,
   "ram_mb": 19256
  },
  {
   "model": "pplx-decider:flash",
   "questions": "triage",
   "load_ms": 7038.2,
   "first_request_ms": 12926.6,
   "answered_by": "pplx-decider:flash",
   "p50_ms": 5533.6,
   "p90_ms": 5592.1,
   "min_ms": 5397.1,
   "eval_p50_ms": 5532.6,
   "input_tokens": 320,
   "device": "cpu",
   "precision": "F32",
   "digest": "e403f9b1f0ac01129919433e754f66b8718ff76a61055eb841ec4f969173fd40",
   "size_mb": 8689,
   "ram_peak_mb": 8356,
   "ram_mb": 8356
  },
  {
   "model": "strands-decider:2b",
   "questions": "triage",
   "load_ms": 31173.1,
   "first_request_ms": 35954.2,
   "answered_by": "strands-decider:2b",
   "p50_ms": 4461.5,
   "p90_ms": 4507.5,
   "min_ms": 4387.5,
   "eval_p50_ms": 4460.5,
   "input_tokens": 562,
   "device": "cpu",
   "precision": "F32",
   "digest": "9aae566845cbf83cafbe804ba9f486fd856239f3d405d265704cb42385bea63a",
   "size_mb": 4571,
   "ram_peak_mb": 11367,
   "ram_mb": 7015
  }
 ],
 "edition": "2026.10"
}