EuroEval Serbian NLU: leaderboard
Metric: NLU Average Score (%). Source: euroeval.com. 209 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 68.85 |
| 2 | GPT-5.4 Mini (Medium) | 67.51 |
| 3 | GPT-5 | 67.25 |
| 4 | GPT-5.4 Mini (High) | 67.09 |
| 5 | Gemini 3.1 Flash Lite (Preview) | 66.73 |
| 6 | Gemini 2.5 Flash (Thinking) | 66.65 |
| 7 | Gemma 4 31B (IT) | 66.22 |
| 8 | Claude Sonnet 4.5 (Thinking) | 66.04 |
| 9 | GPT-5.2 | 65.32 |
| 10 | Gemini 2.5 Pro | 65.24 |
| 11 | Claude Sonnet 4.6 | 63.4 |
| 12 | GPT-5 Mini | 63.17 |
| 13 | Mistral Small 3.2 | 62.7 |
| 14 | gemma-3-27B-pt | 62.63 |
| 15 | Grok 4.1 Fast (Reasoning) | 62.45 |
Interactive version: theaggregate.ai/benchmark?slug=euroeval-serbian-nlu · How It Works · Data refreshed daily, snapshot 2026-09-05.