{
 "schema": 1,
 "suite": "latency-triage",
 "date": "2026-10-25T18:43:35Z",
 "ollaya": "0.18.0",
 "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 0.18.0.",
 "results": [
  {
   "model": "decision2:kai",
   "questions": "triage",
   "load_ms": 10535.9,
   "first_request_ms": 12426.7,
   "answered_by": "decision2:kai",
   "p50_ms": 1797.8,
   "p90_ms": 1810.3,
   "min_ms": 1747.4,
   "eval_p50_ms": 1796.7,
   "input_tokens": 587,
   "device": "cpu",
   "precision": "F32",
   "digest": "c1f010e05fb8fab53826338931bbd5b5cb7b10ad8410fdf5bf1c344bd6e097f5",
   "size_mb": 1454,
   "ram_peak_mb": 4382,
   "ram_mb": 4382
  },
  {
   "model": "decision2:eos",
   "questions": "triage",
   "load_ms": 14872.7,
   "first_request_ms": 18725.9,
   "answered_by": "decision2:eos",
   "p50_ms": 3225.8,
   "p90_ms": 3241.6,
   "min_ms": 3199,
   "eval_p50_ms": 3224.8,
   "input_tokens": 644,
   "device": "cpu",
   "precision": "F32",
   "digest": "4c8c10ac75b300337e9687b249e879a19287fc4481f23e5c5e98020eb650733b",
   "size_mb": 1952,
   "ram_peak_mb": 4704,
   "ram_mb": 3368
  },
  {
   "model": "decision2:sol",
   "questions": "triage",
   "load_ms": 44636.2,
   "first_request_ms": 49612.5,
   "answered_by": "decision2:sol",
   "p50_ms": 4636.9,
   "p90_ms": 4741.8,
   "min_ms": 4379.2,
   "eval_p50_ms": 4635.9,
   "input_tokens": 644,
   "device": "cpu",
   "precision": "F32",
   "digest": "85d3edf9e8bcf7cb39f0907c504781fd2f57a6c0725a24aa4f202dbe103f28a9",
   "size_mb": 4596,
   "ram_peak_mb": 11419,
   "ram_mb": 6642
  },
  {
   "model": "decision2:nox",
   "questions": "triage",
   "load_ms": 7688.1,
   "first_request_ms": 20393.6,
   "answered_by": "decision2:nox",
   "p50_ms": 12130,
   "p90_ms": 12163.5,
   "min_ms": 12082.4,
   "eval_p50_ms": 12129,
   "input_tokens": 644,
   "device": "cpu",
   "precision": "F32",
   "digest": "d81d2a4fe5cad9b1e0e6a8c44d9108732a129328a9e595c935c4b6083b5e9133",
   "size_mb": 9277,
   "ram_peak_mb": 9906,
   "ram_mb": 9906
  },
  {
   "model": "decision2:lux",
   "questions": "triage",
   "load_ms": 10878.6,
   "first_request_ms": 31278.9,
   "answered_by": "decision2:lux",
   "p50_ms": 18670.3,
   "p90_ms": 18728.1,
   "min_ms": 18414.6,
   "eval_p50_ms": 18669.1,
   "input_tokens": 644,
   "device": "cpu",
   "precision": "F32",
   "digest": "4148abeac890bfcf50bab4fdae636336468b21a9291f9a3f42ae526b17fd32a8",
   "size_mb": 17127,
   "ram_peak_mb": 16874,
   "ram_mb": 16874
  }
 ],
 "edition": "2026.10"
}