EuroEval Serbian NLU - ScaLA SR: leaderboard
Metric: Linguistic acceptability Score (%). Source: euroeval.com. 205 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 81.14 |
| 2 | Gemini 3.1 Flash Lite (Preview) | 79.48 |
| 3 | GPT-5.2 | 75.24 |
| 4 | Gemini 2.5 Flash (Thinking) | 74.36 |
| 5 | GPT-5.4 Mini (Medium) | 73.04 |
| 6 | GPT-5.4 Mini (High) | 72.8 |
| 7 | GPT-5 | 72.57 |
| 8 | Gemini 2.5 Pro | 71.6 |
| 9 | Claude Sonnet 4.6 | 68.97 |
| 10 | Gemma 4 31B (IT) | 68.75 |
| 11 | GPT-4.1 | 67.6 |
| 12 | GPT-5 Mini | 67.09 |
| 13 | Claude Sonnet 4.5 (Thinking) | 66.43 |
| 14 | Gemma 3 27B (IT) | 63.44 |
| 15 | GPT-5.4 Nano (Medium) | 62.14 |
Interactive version: theaggregate.ai/benchmark?slug=euroeval-serbian-nlu-scala-sr · How It Works · Data refreshed daily, snapshot 2026-09-05.