EuroEval Serbian NLU - UNER SR: leaderboard
Metric: Named entity recognition Score (%). Source: euroeval.com. 205 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 Mini (High) | 88.77 |
| 2 | GPT-5.4 Mini (Medium) | 88.75 |
| 3 | Claude Sonnet 4.5 (Thinking) | 87.6 |
| 4 | GPT-5 | 87.13 |
| 5 | Claude Sonnet 4.5 (Non-reasoning) | 85.69 |
| 6 | Qwen 3 4B 2507 (Thinking) | 85.21 |
| 7 | Gemini 2.5 Flash (Thinking) | 84.58 |
| 8 | Gemma 4 31B (IT) | 84.1 |
| 9 | Llama 3.3 70B Instruct | 82.88 |
| 10 | Gemma 4 26B A4B (IT) | 81.73 |
| 11 | GPT-5 Nano | 81.22 |
| 12 | Mistral Small 3.1 | 81.05 |
| 13 | Claude Sonnet 4.6 | 80.97 |
| 14 | Mistral Small 3.2 | 80.67 |
| 15 | Gemini 3 Pro (Preview) | 80.51 |
Interactive version: theaggregate.ai/benchmark?slug=euroeval-serbian-nlu-uner-sr · How It Works · Data refreshed daily, snapshot 2026-09-05.