{
 "schema": 1,
 "suite": "latency-triage",
 "date": "2026-10-25T19:28:31Z",
 "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": "d1-omni:600m",
   "questions": "triage",
   "load_ms": 5798.4,
   "first_request_ms": 6812.2,
   "answered_by": "d1-omni:600m",
   "p50_ms": 960.2,
   "p90_ms": 991.8,
   "min_ms": 938.7,
   "eval_p50_ms": 959.2,
   "input_tokens": 374,
   "device": "cpu",
   "precision": "F32",
   "digest": "53dfa4be304b8ab233617f432ed655050d6d459a48f89dae94a574f75414f139",
   "size_mb": 2248,
   "ram_peak_mb": 1820,
   "ram_mb": 1820
  },
  {
   "model": "gliclass-v4:2b",
   "questions": "triage",
   "load_ms": 51877.3,
   "first_request_ms": 55508.7,
   "answered_by": "gliclass-v4:2b",
   "p50_ms": 3571.2,
   "p90_ms": 3658.2,
   "min_ms": 3560.6,
   "eval_p50_ms": 3570.2,
   "input_tokens": 270,
   "device": "cpu",
   "precision": "F32",
   "digest": "e2575bb26e1362f95d78049e2988ba36973ebb95f308b7d3c94aa277dd05a870",
   "size_mb": 4545,
   "ram_peak_mb": 12580,
   "ram_mb": 7073
  },
  {
   "model": "gliclass-v4:9b",
   "questions": "triage",
   "load_ms": 42838.6,
   "first_request_ms": 62950.2,
   "answered_by": "gliclass-v4:9b",
   "p50_ms": 12976.7,
   "p90_ms": 13205.3,
   "min_ms": 12855.1,
   "eval_p50_ms": 12975.6,
   "input_tokens": 270,
   "device": "cpu",
   "precision": "F32",
   "digest": "0e5bbb07b002e136f66bd57e8dd5bf42613df06dfbb55c80ebd016019e737529",
   "size_mb": 17170,
   "ram_peak_mb": 20320,
   "ram_mb": 20320
  },
  {
   "model": "strands-decider:e2b",
   "questions": "triage",
   "load_ms": 14279.4,
   "first_request_ms": 20036.1,
   "answered_by": "strands-decider:e2b",
   "p50_ms": 5540.7,
   "p90_ms": 5574.1,
   "min_ms": 5393.4,
   "eval_p50_ms": 5539.7,
   "input_tokens": 570,
   "device": "cpu",
   "precision": "F32",
   "digest": "76e87da292235d224abe1f24f19596c5280c9d9eabe4e0ff3278697de088d438",
   "size_mb": 10099,
   "ram_peak_mb": 5617,
   "ram_mb": 5617
  },
  {
   "model": "strands-decider:e4b",
   "questions": "triage",
   "load_ms": 9211.8,
   "first_request_ms": 18600.9,
   "answered_by": "strands-decider:e4b",
   "p50_ms": 9439.6,
   "p90_ms": 9487.8,
   "min_ms": 9207.6,
   "eval_p50_ms": 9438.6,
   "input_tokens": 570,
   "device": "cpu",
   "precision": "F32",
   "digest": "b923eec149f21f2b2cc2a294c9b1da22c52cbbdb72b0f6d044e63054765809c3",
   "size_mb": 15710,
   "ram_peak_mb": 9940,
   "ram_mb": 9940
  },
  {
   "model": "decider:2b",
   "questions": "triage",
   "load_ms": 32793.3,
   "first_request_ms": 36988.3,
   "answered_by": "decider:2b",
   "p50_ms": 3963.2,
   "p90_ms": 3987.4,
   "min_ms": 3946.7,
   "eval_p50_ms": 3962.2,
   "input_tokens": 552,
   "device": "cpu",
   "precision": "F32",
   "digest": "6b81ac65a337f2888699e594e1f325aefe7e9718ce0fba8d6da4a3492aba1982",
   "size_mb": 3616,
   "ram_peak_mb": 10732,
   "ram_mb": 6480
  },
  {
   "model": "decider:4b",
   "questions": "triage",
   "load_ms": 19384.9,
   "first_request_ms": 29665.3,
   "answered_by": "decider:4b",
   "p50_ms": 9763.2,
   "p90_ms": 10123.2,
   "min_ms": 9539.7,
   "eval_p50_ms": 9762.3,
   "input_tokens": 552,
   "device": "cpu",
   "precision": "F32",
   "digest": "b4e1276b55eda06fad7af9536a996fc132a1229bb866755cd6f07778dbe98963",
   "size_mb": 8052,
   "ram_peak_mb": 9543,
   "ram_mb": 9543
  },
  {
   "model": "sura:0.8b",
   "questions": "triage",
   "load_ms": 12616.3,
   "first_request_ms": 14241.8,
   "answered_by": "sura:0.8b",
   "p50_ms": 1654.8,
   "p90_ms": 1824.8,
   "min_ms": 1575.4,
   "eval_p50_ms": 1653.8,
   "input_tokens": 320,
   "device": "cpu",
   "precision": "F32",
   "digest": "473fa350466df82396f2a19152bdc15abaa79a6e57fc99f867c3e0dd280ae18a",
   "size_mb": 1654,
   "ram_peak_mb": 4849,
   "ram_mb": 2616
  }
 ],
 "edition": "2026.10"
}