{
 "schema": 1,
 "suite": "latency-triage",
 "date": "2026-10-01T11:30:01Z",
 "ollaya": "0.8.0",
 "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"
 },
 "note": "",
 "results": [
  {
   "model": "clm:8b",
   "questions": "triage",
   "load_ms": 40074.5,
   "first_request_ms": 57384.1,
   "answered_by": "clm:8b",
   "p50_ms": 4972.3,
   "p90_ms": 5185.6,
   "min_ms": 4528,
   "eval_p50_ms": 4971,
   "input_tokens": 337,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "cygnet:12b",
   "questions": "triage",
   "load_ms": 34286,
   "first_request_ms": 55217.4,
   "answered_by": "cygnet:12b",
   "p50_ms": 20636.1,
   "p90_ms": 20725,
   "min_ms": 20317.2,
   "eval_p50_ms": 20634.6,
   "input_tokens": 746,
   "device": "cpu",
   "precision": "Q8_0"
  },
  {
   "model": "decider:0.8b",
   "questions": "triage",
   "load_ms": 17571.1,
   "first_request_ms": 20849.9,
   "answered_by": "decider:0.8b",
   "p50_ms": 3034.7,
   "p90_ms": 3057.7,
   "min_ms": 2947.8,
   "eval_p50_ms": 3033.7,
   "input_tokens": 552,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "decider:2b",
   "questions": "triage",
   "load_ms": 38819.7,
   "first_request_ms": 43552.1,
   "answered_by": "decider:2b",
   "p50_ms": 4397,
   "p90_ms": 4418.8,
   "min_ms": 4382.3,
   "eval_p50_ms": 4396.1,
   "input_tokens": 584,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "decider:2b-vision",
   "questions": "triage",
   "load_ms": 43953.7,
   "first_request_ms": 47175.1,
   "answered_by": "decider:2b-vision",
   "p50_ms": 3123,
   "p90_ms": 3238.6,
   "min_ms": 2929.2,
   "eval_p50_ms": 3122.1,
   "input_tokens": 361,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "decider:4b",
   "questions": "triage",
   "load_ms": 25123,
   "first_request_ms": 35916.2,
   "answered_by": "decider:4b",
   "p50_ms": 10391.2,
   "p90_ms": 10442.7,
   "min_ms": 10212.3,
   "eval_p50_ms": 10390.2,
   "input_tokens": 560,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "decision:eos",
   "questions": "triage",
   "load_ms": 14807.5,
   "first_request_ms": 18452.2,
   "answered_by": "decision:eos",
   "p50_ms": 3401.7,
   "p90_ms": 3439.6,
   "min_ms": 3205.1,
   "eval_p50_ms": 3400.8,
   "input_tokens": 675,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "gliclass:large",
   "questions": "triage",
   "load_ms": 7839.6,
   "first_request_ms": 9882.4,
   "answered_by": "gliclass:large",
   "p50_ms": 1871.7,
   "p90_ms": 1983.7,
   "min_ms": 1645.9,
   "eval_p50_ms": 1870.8,
   "input_tokens": 273,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "jeb:4b",
   "questions": "triage",
   "load_ms": 10449.3,
   "first_request_ms": 15922.6,
   "answered_by": "jeb:4b",
   "p50_ms": 5295.9,
   "p90_ms": 5399.7,
   "min_ms": 5034.8,
   "eval_p50_ms": 5294.5,
   "input_tokens": 574,
   "device": "cpu",
   "precision": "Q8_0"
  },
  {
   "model": "jeb:9b",
   "questions": "triage",
   "load_ms": 19765.7,
   "first_request_ms": 29536.6,
   "answered_by": "jeb:9b",
   "p50_ms": 9506.4,
   "p90_ms": 9684,
   "min_ms": 8979.5,
   "eval_p50_ms": 9505,
   "input_tokens": 579,
   "device": "cpu",
   "precision": "Q8_0"
  },
  {
   "model": "jeb:27b",
   "questions": "triage",
   "load_ms": 49381.7,
   "first_request_ms": 69189.2,
   "answered_by": "jeb:27b",
   "p50_ms": 19985.3,
   "p90_ms": 20753.4,
   "min_ms": 19467.6,
   "eval_p50_ms": 19984,
   "input_tokens": 564,
   "device": "cpu",
   "precision": "Q4_K - MEDIUM"
  },
  {
   "model": "jeeves:9b",
   "questions": "triage",
   "load_ms": 45573.7,
   "first_request_ms": 66732,
   "answered_by": "jeeves:9b",
   "p50_ms": 20785.4,
   "p90_ms": 20872,
   "min_ms": 20557.8,
   "eval_p50_ms": 20784.3,
   "input_tokens": 503,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "jevk5:4b",
   "questions": "triage",
   "load_ms": 11355.9,
   "first_request_ms": 18755.5,
   "answered_by": "jevk5:4b",
   "p50_ms": 7108.8,
   "p90_ms": 7279.9,
   "min_ms": 6627.1,
   "eval_p50_ms": 7107.6,
   "input_tokens": 762,
   "device": "cpu",
   "precision": "Q8_0"
  },
  {
   "model": "kev:0.8b",
   "questions": "triage",
   "load_ms": 11409.1,
   "first_request_ms": 14413.5,
   "answered_by": "kev:0.8b",
   "p50_ms": 2836,
   "p90_ms": 2854,
   "min_ms": 2829.7,
   "eval_p50_ms": 2835.1,
   "input_tokens": 283,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "kev:4b",
   "questions": "triage",
   "load_ms": 27143,
   "first_request_ms": 35798.9,
   "answered_by": "kev:4b",
   "p50_ms": 8297.8,
   "p90_ms": 8376.8,
   "min_ms": 8000.1,
   "eval_p50_ms": 8296.9,
   "input_tokens": 288,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "kev:9b",
   "questions": "triage",
   "load_ms": 43014.8,
   "first_request_ms": 55722.2,
   "answered_by": "kev:9b",
   "p50_ms": 12116.9,
   "p90_ms": 12210.2,
   "min_ms": 11928.8,
   "eval_p50_ms": 12115.9,
   "input_tokens": 268,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "laya:en",
   "questions": "triage",
   "load_ms": 6559.8,
   "first_request_ms": 7467,
   "answered_by": "laya:en",
   "p50_ms": 915.1,
   "p90_ms": 925.4,
   "min_ms": 906.7,
   "eval_p50_ms": 914.1,
   "input_tokens": 344,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "laya:multilingual",
   "questions": "triage",
   "load_ms": 5772.2,
   "first_request_ms": 6278.8,
   "answered_by": "laya:multilingual",
   "p50_ms": 472.3,
   "p90_ms": 517.2,
   "min_ms": 454.7,
   "eval_p50_ms": 471.5,
   "input_tokens": 346,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "nimble:9b",
   "questions": "triage",
   "load_ms": 46948.7,
   "first_request_ms": 111549.2,
   "answered_by": "nimble:9b",
   "p50_ms": 59778.9,
   "p90_ms": 59996.6,
   "min_ms": 59505.5,
   "eval_p50_ms": 59777.9,
   "input_tokens": 2978,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "nli:deberta-v3-large",
   "questions": "triage",
   "load_ms": 8036.9,
   "first_request_ms": 10482,
   "answered_by": "nli:deberta-v3-large",
   "p50_ms": 2405.2,
   "p90_ms": 2427.3,
   "min_ms": 2052.3,
   "eval_p50_ms": 2404.3,
   "input_tokens": 593,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "nli:modernbert-large",
   "questions": "triage",
   "load_ms": 6115.4,
   "first_request_ms": 7829.5,
   "answered_by": "nli:modernbert-large",
   "p50_ms": 1848.3,
   "p90_ms": 1909,
   "min_ms": 1657,
   "eval_p50_ms": 1847.4,
   "input_tokens": 382,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "qwen3guard:0.6b",
   "questions": "builtin (4)",
   "load_ms": 0,
   "first_request_ms": 1440.7,
   "answered_by": "qwen3guard:0.6b",
   "p50_ms": 1441.6,
   "p90_ms": 1550.7,
   "min_ms": 1421.9,
   "eval_p50_ms": 1440.7,
   "input_tokens": 638,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "von:1.1",
   "questions": "triage",
   "load_ms": 6692.7,
   "first_request_ms": 9254.3,
   "answered_by": "von:1.1",
   "p50_ms": 2445.2,
   "p90_ms": 2512.4,
   "min_ms": 2173.4,
   "eval_p50_ms": 2444.3,
   "input_tokens": 379,
   "device": "cpu",
   "precision": "F32"
  },
  {
   "model": "winnow:e4b",
   "questions": "triage",
   "load_ms": 16187.6,
   "first_request_ms": 20179.9,
   "answered_by": "winnow:e4b",
   "p50_ms": 3992.9,
   "p90_ms": 4036,
   "min_ms": 3714.3,
   "eval_p50_ms": 3991.7,
   "input_tokens": 691,
   "device": "cpu",
   "precision": "Q8_0"
  },
  {
   "model": "winnow:12b",
   "questions": "triage",
   "load_ms": 27874.7,
   "first_request_ms": 38900.4,
   "answered_by": "winnow:12b",
   "p50_ms": 10927.7,
   "p90_ms": 11024.8,
   "min_ms": 10869.7,
   "eval_p50_ms": 10926.7,
   "input_tokens": 711,
   "device": "cpu",
   "precision": "Q8_0"
  }
 ]
}
