EuroEval Latvian NLU: leaderboard
Metric: NLU Average Score (%). Source: euroeval.com. 238 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 61.73 |
| 2 | Gemini 3 Pro (Preview) | 59.53 |
| 3 | GPT-5.4 Mini (High) | 59.35 |
| 4 | GPT-5.4 Mini (Medium) | 58.94 |
| 5 | Gemini 2.5 Flash (Thinking) | 58.37 |
| 6 | Gemini 3.1 Flash Lite (Preview) | 58.1 |
| 7 | Claude Sonnet 4.5 (Thinking) | 57.39 |
| 8 | GPT-5.2 | 56.89 |
| 9 | Gemma 4 31B (IT) | 56.85 |
| 10 | Claude Sonnet 4.6 | 55.87 |
| 11 | Qwen 3 235B A22B 2507 Instruct | 55.56 |
| 12 | Llama 3.1 70B | 53.95 |
| 13 | Mistral Small 3.2 | 53.94 |
| 14 | Qwen 2.5 72B Instruct | 53.83 |
| 15 | Gemini 2.5 Pro | 53.75 |
Interactive version: theaggregate.ai/benchmark?slug=euroeval-latvian-nlu · How It Works · Data refreshed daily, snapshot 2026-09-05.