LingOly: leaderboard
Linguistics-olympiad benchmark testing few-shot reasoning over low-resource and constructed languages from puzzle-style problems.
Metric: Exact Match Accuracy. Source: huggingface.co. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3 Opus | 46.3 | #571 |
| 2 | GPT-4o | 37.6 | #343 |
| 3 | GPT-4 | 33.4 | #438 |
| 4 | Gemini 1.5 Pro | 32.1 | #445 |
| 5 | Command-R+ | 21.5 | #891 |
| 6 | GPT-3.5 | 21.2 | #818 |
| 7 | Mixtral 8x7B | 14.2 | #1143 |
| 8 | Llama 3 8B | 11.4 | #1242 |
| 9 | Llama 3 70B | 10.3 | #631 |
| 10 | Llama 2 70B | 6.4 | #1201 |
| 11 | gemma-7B | 4.9 | #1384 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=lingoly · How It Works · Data refreshed daily, snapshot 2026-10-20.