GaelEval - Cultural Knowledge (Gaelic Prompt): leaderboard
GaelEval (2026) is a Scottish Gaelic benchmark from the DR-LIB project. This board is accuracy in percent on 1,087 four-option questions about Gaelic culture that an LLM generated from transcripts of the An Litir Bheag learner podcast and filtered for answerability, asked with a Gaelic-language prompt; higher is better.
Metric: Accuracy (%) on the 1087-question Gaelic cultural-knowledge MCQ with a Gaelic-language prompt, four options; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 91.35 | #152 |
| 2 | Gemini 3 Pro (Preview) | 91.17 | #119 |
| 3 | GPT-5 | 85.28 | #178 |
| 4 | Gemini 2.5 Flash | 79.85 | #463 |
| 5 | Claude Opus 4.6 | 79.48 | #113 |
| 6 | GPT-5 Mini | 71.48 | #364 |
| 7 | GPT-5.2 | 66.7 | #185 |
| 8 | GPT-4.1 | 66.24 | #505 |
| 9 | GPT-4o | 63.85 | #751 |
| 10 | GLM-4.7 | 63.75 | #380 |
| 11 | Llama 4 Maverick | 59.06 | #1110 |
| 12 | DeepSeek R1 | 58.33 | #482 |
| 13 | GPT-5 Nano | 54.37 | #906 |
| 14 | GPT-OSS-120B | 52.81 | #704 |
| 15 | GPT-4.1 Mini | 47.93 | #789 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=gaeleval-cultural-knowledge-gaelic-prompt · How It Works · Data refreshed daily, snapshot 2026-11-21.