{
  "schema_version": "2.0.0",
  "benchmark": "teanga-bench",
  "benchmark_name": "Teanga Bench",
  "benchmark_version": "0.2.0",
  "suite": "blas",
  "generated_at": "2026-11-15T23:00:41+00:00",
  "note": "Only live runs with status complete are ranked. Scores are means over items with a 95% bootstrap interval; the suite score weights every test equally. cost_usd is the spend on this suite's tests. by_language gives the mean of per-test scores.",
  "models": [
    {
      "rank": 1,
      "ranked": true,
      "model": "openai/gpt-6-luna",
      "family": "openai",
      "page": "models/openai-gpt-6-luna.html",
      "run_file": "openai__gpt-6-luna.json",
      "run_kind": "live",
      "status": "complete",
      "generated_at": "2026-11-15T23:00:41+00:00",
      "data": "full",
      "epochs": 3,
      "overall": {
        "score": 0.7557093823749871,
        "ci95": [
          0.7390800813313367,
          0.7719872330097396
        ],
        "n_tests": 5,
        "n_items": 1500
      },
      "cost_usd": 0.775096764,
      "cost_per_correct_usd": 0.00022377588392287742,
      "by_language": {
        "breton": 0.7260338915215823,
        "cornish": 0.7004943837532529,
        "irish": 0.81119120560192,
        "manx": 0.6693204619241256,
        "scots-gaelic": 0.783916552898724,
        "welsh": 0.8346917682123323
      },
      "tests": {
        "lesson_exercises": {
          "score": 0.9361111111111111,
          "ci95": [
            0.9111111111111111,
            0.9574074074074073
          ],
          "n_items": 360,
          "epochs": 3,
          "consistency": 0.9611111111111111,
          "status": "complete",
          "dataset": "full",
          "cost_usd": 0.0545398,
          "cost_per_correct_usd": 5.394638971315529e-05
        },
        "morphology": {
          "score": 0.8046296296296297,
          "ci95": [
            0.7685185185185186,
            0.8407407407407407
          ],
          "n_items": 360,
          "epochs": 3,
          "consistency": 0.8555555555555555,
          "status": "complete",
          "dataset": "full",
          "cost_usd": 0.1706622,
          "cost_per_correct_usd": 0.0001963891829689298
        },
        "round_trip": {
          "score": 0.6447548341165746,
          "ci95": [
            0.6171525369055623,
            0.6740683205592297
          ],
          "n_items": 240,
          "epochs": 3,
          "consistency": 0.32916666666666666,
          "status": "complete",
          "dataset": "full",
          "cost_usd": 0.1443668,
          "cost_per_correct_usd": 0.0003109855620069556
        },
        "translation": {
          "score": 0.6800883740546572,
          "ci95": [
            0.6536907211785644,
            0.7077288342979027
          ],
          "n_items": 360,
          "epochs": 3,
          "consistency": 0.4388888888888889,
          "status": "complete",
          "dataset": "full",
          "cost_usd": 0.1218164,
          "cost_per_correct_usd": 0.00016585045012679198
        },
        "tutor_accuracy": {
          "score": 0.7129629629629629,
          "ci95": [
            0.6537037037037037,
            0.7703703703703701
          ],
          "n_items": 180,
          "epochs": 3,
          "consistency": 0.7944444444444444,
          "status": "complete",
          "dataset": "full",
          "cost_usd": 0.283711564,
          "cost_per_correct_usd": 0.0007369131532467535
        }
      }
    }
  ]
}