{
 "schema": 1,
 "date": "2026-10-23",
 "edition": "2026.10",
 "machine": "rtx-5090",
 "device": "cuda",
 "suite": "public-benchmark",
 "servers": {
  "ollaya": "0.17.0"
 },
 "protocol": {
  "benchmark": "Bespoke Labs' public decision benchmark: 13 subsets of 11 human-labeled public datasets, 3,880 questions",
  "source": "https://github.com/bespokelabsai/nimble/blob/main/docs/PUBLIC_BENCHMARKS.md",
  "runner": "Bespoke's own evaluate_public_jev, through convert/ollaya_convert/bench_public.py",
  "client": "one request (one question) at a time over HTTP to /v1/systemone, timed by the client",
  "accuracy": "macro: the mean of the 13 subsets; a rejected request counts as wrong",
  "ece": "ten-bin expected calibration error of the top probability; pooled over all 3,880 questions"
 },
 "note": "Ollaya Bench edition 2026.10.",
 "results": [
  {
   "server": "ollaya",
   "model": "coba:400m-ternary",
   "digest": "ec26808146ae191e76ec72691bd0178c2fb4804407436df68575af7fcdcee7f8",
   "records": 3880,
   "errors": 453,
   "macro_accuracy": 0.6537,
   "macro_accuracy_ci": [
    0.6422,
    0.6647
   ],
   "micro_accuracy": 0.7023,
   "macro_ece": 0.0847,
   "pooled_ece": 0.0466,
   "brier": 0.2841,
   "median_ms": 20.2,
   "p95_ms": 24.5,
   "by_type": {
    "choice": 0.7542,
    "noul": 0.8688,
    "score": 0.1276
   },
   "per_subset": {
    "vitaminc-dev": {
     "n": 599,
     "errors": 0,
     "accuracy": 0.8948,
     "ece": 0.0242
    },
    "massive-en-US": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.9,
     "ece": 0.0346
    },
    "massive-de-DE": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.4171,
     "ece": 0.1988
    },
    "boolq": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.81,
     "ece": 0.0933
    },
    "squad2": {
     "n": 299,
     "errors": 19,
     "accuracy": 0.8395,
     "ece": 0.0608
    },
    "paws": {
     "n": 250,
     "errors": 0,
     "accuracy": 0.936,
     "ece": 0.0303
    },
    "multinli": {
     "n": 299,
     "errors": 0,
     "accuracy": 0.903,
     "ece": 0.0387
    },
    "civil_comments": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.9267,
     "ece": 0.0284
    },
    "aegis2": {
     "n": 250,
     "errors": 8,
     "accuracy": 0.832,
     "ece": 0.0634
    },
    "helpsteer2": {
     "n": 249,
     "errors": 144,
     "accuracy": 0.1606,
     "ece": 0.0795
    },
    "summeval-relevance": {
     "n": 240,
     "errors": 170,
     "accuracy": 0.0833,
     "ece": 0.1587
    },
    "summeval-consistency": {
     "n": 144,
     "errors": 101,
     "accuracy": 0.1389,
     "ece": 0.1621
    },
    "pubmedqa": {
     "n": 250,
     "errors": 11,
     "accuracy": 0.656,
     "ece": 0.1287
    }
   }
  },
  {
   "server": "ollaya",
   "model": "coba:150m-ternary",
   "digest": "e1a3fad29866d5b4413179e84f3945d1b7fd79e5a9faa6d6f5e131ed8d18c020",
   "records": 3880,
   "errors": 453,
   "macro_accuracy": 0.6222,
   "macro_accuracy_ci": [
    0.6098,
    0.6345
   ],
   "micro_accuracy": 0.6634,
   "macro_ece": 0.0956,
   "pooled_ece": 0.0526,
   "brier": 0.3361,
   "median_ms": 17.4,
   "p95_ms": 21.1,
   "by_type": {
    "choice": 0.6986,
    "noul": 0.835,
    "score": 0.1401
   },
   "per_subset": {
    "vitaminc-dev": {
     "n": 599,
     "errors": 0,
     "accuracy": 0.828,
     "ece": 0.0495
    },
    "massive-en-US": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.88,
     "ece": 0.0481
    },
    "massive-de-DE": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.3429,
     "ece": 0.2216
    },
    "boolq": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.73,
     "ece": 0.0928
    },
    "squad2": {
     "n": 299,
     "errors": 19,
     "accuracy": 0.7726,
     "ece": 0.0758
    },
    "paws": {
     "n": 250,
     "errors": 0,
     "accuracy": 0.928,
     "ece": 0.0323
    },
    "multinli": {
     "n": 299,
     "errors": 0,
     "accuracy": 0.8261,
     "ece": 0.0272
    },
    "civil_comments": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.9167,
     "ece": 0.0385
    },
    "aegis2": {
     "n": 250,
     "errors": 8,
     "accuracy": 0.828,
     "ece": 0.089
    },
    "helpsteer2": {
     "n": 249,
     "errors": 144,
     "accuracy": 0.1647,
     "ece": 0.0637
    },
    "summeval-relevance": {
     "n": 240,
     "errors": 170,
     "accuracy": 0.075,
     "ece": 0.1831
    },
    "summeval-consistency": {
     "n": 144,
     "errors": 101,
     "accuracy": 0.1806,
     "ece": 0.1761
    },
    "pubmedqa": {
     "n": 250,
     "errors": 11,
     "accuracy": 0.616,
     "ece": 0.1451
    }
   }
  },
  {
   "server": "ollaya",
   "model": "coba:68m-ternary",
   "digest": "93063c00de80a702a38b1acdd28747ae353fe49a569e3f0bdac61896c9b4d3fa",
   "records": 3880,
   "errors": 453,
   "macro_accuracy": 0.6033,
   "macro_accuracy_ci": [
    0.5901,
    0.6166
   ],
   "micro_accuracy": 0.6436,
   "macro_ece": 0.1109,
   "pooled_ece": 0.0765,
   "brier": 0.3809,
   "median_ms": 15.4,
   "p95_ms": 19.3,
   "by_type": {
    "choice": 0.6859,
    "noul": 0.8005,
    "score": 0.1368
   },
   "per_subset": {
    "vitaminc-dev": {
     "n": 599,
     "errors": 0,
     "accuracy": 0.7963,
     "ece": 0.0191
    },
    "massive-en-US": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.8943,
     "ece": 0.0373
    },
    "massive-de-DE": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.3114,
     "ece": 0.4141
    },
    "boolq": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.6967,
     "ece": 0.099
    },
    "squad2": {
     "n": 299,
     "errors": 19,
     "accuracy": 0.7191,
     "ece": 0.0953
    },
    "paws": {
     "n": 250,
     "errors": 0,
     "accuracy": 0.892,
     "ece": 0.0637
    },
    "multinli": {
     "n": 299,
     "errors": 0,
     "accuracy": 0.8194,
     "ece": 0.0364
    },
    "civil_comments": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.9267,
     "ece": 0.0312
    },
    "aegis2": {
     "n": 250,
     "errors": 8,
     "accuracy": 0.768,
     "ece": 0.0774
    },
    "helpsteer2": {
     "n": 249,
     "errors": 144,
     "accuracy": 0.1647,
     "ece": 0.0605
    },
    "summeval-relevance": {
     "n": 240,
     "errors": 170,
     "accuracy": 0.0792,
     "ece": 0.1091
    },
    "summeval-consistency": {
     "n": 144,
     "errors": 101,
     "accuracy": 0.1667,
     "ece": 0.2276
    },
    "pubmedqa": {
     "n": 250,
     "errors": 11,
     "accuracy": 0.608,
     "ece": 0.1715
    }
   }
  },
  {
   "server": "ollaya",
   "model": "sura:0.8b",
   "digest": "b5a8e8f48b6aeba1bb805c5941d992832458e8e5e868fd8f80302915ff063f7f",
   "records": 3880,
   "errors": 0,
   "macro_accuracy": 0.5861,
   "macro_accuracy_ci": [
    0.5699,
    0.6013
   ],
   "micro_accuracy": 0.6088,
   "macro_ece": 0.1324,
   "pooled_ece": 0.0573,
   "brier": 0.509,
   "median_ms": 59.3,
   "p95_ms": 80.3,
   "by_type": {
    "choice": 0.6396,
    "noul": 0.6672,
    "score": 0.3619
   },
   "per_subset": {
    "vitaminc-dev": {
     "n": 599,
     "errors": 0,
     "accuracy": 0.7212,
     "ece": 0.0382
    },
    "massive-en-US": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.7514,
     "ece": 0.0299
    },
    "massive-de-DE": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.6171,
     "ece": 0.066
    },
    "boolq": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.6967,
     "ece": 0.0495
    },
    "squad2": {
     "n": 299,
     "errors": 0,
     "accuracy": 0.6087,
     "ece": 0.214
    },
    "paws": {
     "n": 250,
     "errors": 0,
     "accuracy": 0.664,
     "ece": 0.0657
    },
    "multinli": {
     "n": 299,
     "errors": 0,
     "accuracy": 0.6321,
     "ece": 0.071
    },
    "civil_comments": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.8467,
     "ece": 0.0449
    },
    "aegis2": {
     "n": 250,
     "errors": 0,
     "accuracy": 0.52,
     "ece": 0.2325
    },
    "helpsteer2": {
     "n": 249,
     "errors": 0,
     "accuracy": 0.3052,
     "ece": 0.158
    },
    "summeval-relevance": {
     "n": 240,
     "errors": 0,
     "accuracy": 0.1417,
     "ece": 0.4137
    },
    "summeval-consistency": {
     "n": 144,
     "errors": 0,
     "accuracy": 0.6389,
     "ece": 0.262
    },
    "pubmedqa": {
     "n": 250,
     "errors": 0,
     "accuracy": 0.476,
     "ece": 0.0753
    }
   }
  },
  {
   "server": "ollaya",
   "model": "coba:1b",
   "digest": "7aaf4eaed0da754d7766714cd5b661c33f5f22476001f7e8fb0bd11f12eeacde",
   "records": 3880,
   "errors": 453,
   "macro_accuracy": 0.561,
   "macro_accuracy_ci": [
    0.5476,
    0.5743
   ],
   "micro_accuracy": 0.6046,
   "macro_ece": 0.1594,
   "pooled_ece": 0.0903,
   "brier": 0.4488,
   "median_ms": 21.4,
   "p95_ms": 25,
   "by_type": {
    "choice": 0.6799,
    "noul": 0.701,
    "score": 0.1297
   },
   "per_subset": {
    "vitaminc-dev": {
     "n": 599,
     "errors": 0,
     "accuracy": 0.7513,
     "ece": 0.0638
    },
    "massive-en-US": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.7743,
     "ece": 0.0781
    },
    "massive-de-DE": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.5114,
     "ece": 0.1601
    },
    "boolq": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.7567,
     "ece": 0.1005
    },
    "squad2": {
     "n": 299,
     "errors": 19,
     "accuracy": 0.6689,
     "ece": 0.0938
    },
    "paws": {
     "n": 250,
     "errors": 0,
     "accuracy": 0.6,
     "ece": 0.2509
    },
    "multinli": {
     "n": 299,
     "errors": 0,
     "accuracy": 0.7425,
     "ece": 0.0493
    },
    "civil_comments": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.8833,
     "ece": 0.0342
    },
    "aegis2": {
     "n": 250,
     "errors": 8,
     "accuracy": 0.596,
     "ece": 0.2231
    },
    "helpsteer2": {
     "n": 249,
     "errors": 144,
     "accuracy": 0.1446,
     "ece": 0.4065
    },
    "summeval-relevance": {
     "n": 240,
     "errors": 170,
     "accuracy": 0.0708,
     "ece": 0.3184
    },
    "summeval-consistency": {
     "n": 144,
     "errors": 101,
     "accuracy": 0.1736,
     "ece": 0.1898
    },
    "pubmedqa": {
     "n": 250,
     "errors": 11,
     "accuracy": 0.62,
     "ece": 0.1034
    }
   }
  },
  {
   "server": "ollaya",
   "model": "coba:400m",
   "digest": "99c4f5b9e23aef32e9ea990a192a6990f5679a9b095ffde4dd7889839405b9e6",
   "records": 3880,
   "errors": 453,
   "macro_accuracy": 0.5606,
   "macro_accuracy_ci": [
    0.5465,
    0.5745
   ],
   "micro_accuracy": 0.6054,
   "macro_ece": 0.1476,
   "pooled_ece": 0.0892,
   "brier": 0.4408,
   "median_ms": 21.2,
   "p95_ms": 24.8,
   "by_type": {
    "choice": 0.6898,
    "noul": 0.6967,
    "score": 0.1187
   },
   "per_subset": {
    "vitaminc-dev": {
     "n": 599,
     "errors": 0,
     "accuracy": 0.7245,
     "ece": 0.0719
    },
    "massive-en-US": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.8,
     "ece": 0.0402
    },
    "massive-de-DE": {
     "n": 350,
     "errors": 0,
     "accuracy": 0.5657,
     "ece": 0.1169
    },
    "boolq": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.7467,
     "ece": 0.0848
    },
    "squad2": {
     "n": 299,
     "errors": 19,
     "accuracy": 0.6856,
     "ece": 0.1138
    },
    "paws": {
     "n": 250,
     "errors": 0,
     "accuracy": 0.624,
     "ece": 0.185
    },
    "multinli": {
     "n": 299,
     "errors": 0,
     "accuracy": 0.7625,
     "ece": 0.0428
    },
    "civil_comments": {
     "n": 300,
     "errors": 0,
     "accuracy": 0.8833,
     "ece": 0.0292
    },
    "aegis2": {
     "n": 250,
     "errors": 8,
     "accuracy": 0.544,
     "ece": 0.2822
    },
    "helpsteer2": {
     "n": 249,
     "errors": 144,
     "accuracy": 0.1365,
     "ece": 0.369
    },
    "summeval-relevance": {
     "n": 240,
     "errors": 170,
     "accuracy": 0.0875,
     "ece": 0.318
    },
    "summeval-consistency": {
     "n": 144,
     "errors": 101,
     "accuracy": 0.1319,
     "ece": 0.1662
    },
    "pubmedqa": {
     "n": 250,
     "errors": 11,
     "accuracy": 0.596,
     "ece": 0.0992
    }
   }
  }
 ]
}
