{
 "schema": 1,
 "suite": "latency-triage",
 "date": "2026-10-01T11:18:36Z",
 "ollaya": "0.8.0",
 "machine": "rtx-4090",
 "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": "",
 "results": [
  {
   "model": "cygnet:12b",
   "questions": "triage",
   "load_ms": 2009.4,
   "first_request_ms": 2207.6,
   "answered_by": "cygnet:12b",
   "p50_ms": 196.6,
   "p90_ms": 197.9,
   "min_ms": 192.8,
   "eval_p50_ms": 196,
   "input_tokens": 746,
   "device": "cuda:0",
   "precision": "Q8_0"
  },
  {
   "model": "decider:0.8b",
   "questions": "triage",
   "load_ms": 6279.8,
   "first_request_ms": 6487.6,
   "answered_by": "decider:0.8b",
   "p50_ms": 169.4,
   "p90_ms": 174.8,
   "min_ms": 165.6,
   "eval_p50_ms": 168.8,
   "input_tokens": 552,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "decider:2b",
   "questions": "triage",
   "load_ms": 11724.3,
   "first_request_ms": 11979.3,
   "answered_by": "decider:2b",
   "p50_ms": 216.2,
   "p90_ms": 219,
   "min_ms": 212.9,
   "eval_p50_ms": 215.5,
   "input_tokens": 552,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "decider:2b-vision",
   "questions": "triage",
   "load_ms": 14031.6,
   "first_request_ms": 14246.2,
   "answered_by": "decider:2b-vision",
   "p50_ms": 142.3,
   "p90_ms": 143.5,
   "min_ms": 139.8,
   "eval_p50_ms": 141.7,
   "input_tokens": 356,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "decider:4b",
   "questions": "triage",
   "load_ms": 5614.2,
   "first_request_ms": 6186.3,
   "answered_by": "decider:4b",
   "p50_ms": 514.3,
   "p90_ms": 543.5,
   "min_ms": 511.6,
   "eval_p50_ms": 513.7,
   "input_tokens": 560,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "decision:eos",
   "questions": "triage",
   "load_ms": 6284.2,
   "first_request_ms": 6530.7,
   "answered_by": "decision:eos",
   "p50_ms": 209.9,
   "p90_ms": 217.2,
   "min_ms": 206.2,
   "eval_p50_ms": 209.3,
   "input_tokens": 675,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "kev:4b",
   "questions": "triage",
   "load_ms": 5571.2,
   "first_request_ms": 5963.4,
   "answered_by": "kev:4b",
   "p50_ms": 360.2,
   "p90_ms": 362.3,
   "min_ms": 356.6,
   "eval_p50_ms": 359.6,
   "input_tokens": 298,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "kev:9b",
   "questions": "triage",
   "load_ms": 9019.9,
   "first_request_ms": 9571.7,
   "answered_by": "kev:9b",
   "p50_ms": 506.2,
   "p90_ms": 508.1,
   "min_ms": 502.3,
   "eval_p50_ms": 505.6,
   "input_tokens": 288,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "laya:en",
   "questions": "triage",
   "load_ms": 1167.7,
   "first_request_ms": 1179.5,
   "answered_by": "laya:en",
   "p50_ms": 10.6,
   "p90_ms": 10.8,
   "min_ms": 10.2,
   "eval_p50_ms": 10.1,
   "input_tokens": 344,
   "device": "cuda:0",
   "precision": "F16"
  },
  {
   "model": "laya:multilingual",
   "questions": "triage",
   "load_ms": 1291.3,
   "first_request_ms": 1301,
   "answered_by": "laya:multilingual",
   "p50_ms": 7.4,
   "p90_ms": 8.2,
   "min_ms": 7.1,
   "eval_p50_ms": 6.9,
   "input_tokens": 356,
   "device": "cuda:0",
   "precision": "F16"
  },
  {
   "model": "winnow:12b",
   "questions": "triage",
   "load_ms": 5089,
   "first_request_ms": 5246.2,
   "answered_by": "winnow:12b",
   "p50_ms": 157.6,
   "p90_ms": 158.4,
   "min_ms": 156.2,
   "eval_p50_ms": 157,
   "input_tokens": 696,
   "device": "cuda:0",
   "precision": "Q8_0"
  },
  {
   "model": "winnow:e4b",
   "questions": "triage",
   "load_ms": 3064.3,
   "first_request_ms": 3174.6,
   "answered_by": "winnow:e4b",
   "p50_ms": 103,
   "p90_ms": 104,
   "min_ms": 101.6,
   "eval_p50_ms": 102.3,
   "input_tokens": 691,
   "device": "cuda:0",
   "precision": "Q8_0"
  },
  {
   "model": "gliclass:latest",
   "questions": "triage",
   "load_ms": 2341.6,
   "first_request_ms": 2359.5,
   "answered_by": "gliclass:latest",
   "p50_ms": 16.9,
   "p90_ms": 17.1,
   "min_ms": 16.5,
   "eval_p50_ms": 16.2,
   "input_tokens": 258,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "jevk5:latest",
   "questions": "triage",
   "load_ms": 1899.3,
   "first_request_ms": 2002,
   "answered_by": "jevk5:latest",
   "p50_ms": 102.6,
   "p90_ms": 103.4,
   "min_ms": 102.2,
   "eval_p50_ms": 102,
   "input_tokens": 752,
   "device": "cuda:0",
   "precision": "Q8_0"
  },
  {
   "model": "nli:latest",
   "questions": "triage",
   "load_ms": 2648.9,
   "first_request_ms": 2672.3,
   "answered_by": "nli:latest",
   "p50_ms": 22.4,
   "p90_ms": 23.2,
   "min_ms": 21.7,
   "eval_p50_ms": 21.7,
   "input_tokens": 567,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "qwen3guard:latest",
   "questions": "builtin (4)",
   "load_ms": 0,
   "first_request_ms": 32.5,
   "answered_by": "qwen3guard:latest",
   "p50_ms": 33.6,
   "p90_ms": 34,
   "min_ms": 30.3,
   "eval_p50_ms": 33,
   "input_tokens": 642,
   "device": "cuda:0",
   "precision": "F32"
  },
  {
   "model": "von:latest",
   "questions": "triage",
   "load_ms": 1690.9,
   "first_request_ms": 1712.1,
   "answered_by": "von:latest",
   "p50_ms": 22.5,
   "p90_ms": 22.8,
   "min_ms": 21.1,
   "eval_p50_ms": 21.9,
   "input_tokens": 394,
   "device": "cuda:0",
   "precision": "F32"
  }
 ],
 "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"
 }
}
