EuroEval Serbian NLU: leaderboard
Serbian-language EuroEval natural-language-understanding suite, separating NLU task performance from the broader all-task EuroEval aggregate.
Metric: NLU Average Score (%). Source: euroeval.com. Saturation forecast: Estimated already saturated. 209 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 68.85 | #66 |
| 2 | GPT-5 | 67.25 | #98 |
| 3 | GPT-5.4 Mini (2026-03-17) (High) | 67.09 | #169 (GPT-5.4 Mini (2026-03-17)) |
| 4 | Gemini 3 Flash (Preview) (Thinking) | 66.8 | #82 (Gemini 3 Flash (Preview)) |
| 5 | Gemini 3.1 Flash Lite (Preview) | 66.73 | #194 |
| 6 | Gemini 2.5 Flash (Thinking) | 66.65 | #245 (Gemini 2.5 Flash) |
| 7 | Gemini 3 Flash (Preview) (Non-reasoning) | 66.44 | #82 (Gemini 3 Flash (Preview)) |
| 8 | Gemma 4 31B (IT) | 66.22 | #176 |
| 9 | Claude Sonnet 4.5 (Thinking) | 66.04 | #142 (Claude Sonnet 4.5) |
| 10 | GPT-5.2 | 65.32 | #110 |
| 11 | Gemini 2.5 Pro | 65.24 | #152 |
| 12 | Qwen 3.5 35B A3B Base | 64.28 | #310 |
| 13 | Claude Sonnet 4.6 | 63.4 | #90 |
| 14 | GPT-5 Mini (2025-08-07) | 63.17 | #172 |
| 15 | Mistral Small 3.2 | 62.7 | #605 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-serbian-nlu · How It Works · Data refreshed daily, snapshot 2026-10-20.