GaelEval - Gaelic to English Translation (BLEU): leaderboard
GaelEval (2026) is a Scottish Gaelic benchmark from the DR-LIB project. This board is corpus BLEU (0 to 100) for translating 908 An Litir Bheag podcast transcripts from Gaelic into English, scored against the professional parallel translations; higher is better.
Metric: BLEU-4 of Gaelic-to-English translation of An Litir Bheag transcripts against gold references; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 75.28 | #482 |
| 2 | Gemini 2.5 Flash | 74.36 | #463 |
| 3 | Llama 4 Maverick | 73.49 | #1110 |
| 4 | Gemini 3 Pro (Preview) | 73.41 | #119 |
| 5 | Claude Haiku 4.5 | 72.32 | #583 |
| 6 | Gemini 3 Flash (Preview) | 71.75 | #152 |
| 7 | GPT-5.2 | 71.02 | #185 |
| 8 | Claude Opus 4.6 | 70.37 | #113 |
| 9 | GPT-5 | 70.27 | #178 |
| 10 | GPT-5 Mini | 69.58 | #364 |
| 11 | GPT-4.1 Mini | 69.09 | #789 |
| 12 | GPT-4o Mini | 69.01 | #1667 |
| 13 | GPT-4o | 68.73 | #751 |
| 14 | GPT-4.1 | 66.6 | #505 |
| 15 | GPT-4.1 Nano | 63.31 | #2056 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=gaeleval-gaelic-to-english-translation-bleu · How It Works · Data refreshed daily, snapshot 2026-11-21.