EuroEval Romanian NLU - RoSent: leaderboard
The "RoSent" subtask of EuroEval Romanian NLU. Romanian-language EuroEval natural-language-understanding suite, separating NLU task performance from the broader all-task EuroEval aggregate.
Metric: Matthews correlation coefficient (%). Source: euroeval.com. Saturation forecast: Estimated already saturated. 207 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini (2025-08-07) | 98.51 | #172 |
| 2 | Llama 3.3 70B Instruct | 98.47 | #534 |
| 3 | Gemma 3 27B (IT) | 97.74 | #523 |
| 4 | Qwen 2.5 72B Instruct | 97.62 | #454 |
| 5 | Claude Sonnet 4.5 (Non-reasoning) | 97.6 | #142 (Claude Sonnet 4.5) |
| 6 | Mistral Small 3.1 | 97.56 | #610 |
| 7 | Qwen 3 32B (Non-reasoning) | 97.53 | #443 (Qwen 3 32B) |
| 8 | Grok 4.1 Fast (Reasoning) | 97.5 | #217 (Grok 4.1 Fast) |
| 9 | cogito-v1-preview-qwen-14B | 97.31 | |
| 10 | Gemma 3 12B (IT) | 97.15 | #664 |
| 11 | GPT-5 | 97.1 | #98 |
| 12 | Qwen 3 14B (Non-reasoning) | 96.98 | #537 (Qwen 3 14B) |
| 13 | Mistral Small 3.2 | 96.97 | #605 |
| 14 | Gemini 3 Flash (Preview) (Non-reasoning) | 96.82 | #82 (Gemini 3 Flash (Preview)) |
| 15 | Claude Sonnet 4.5 (Thinking) | 96.8 | #142 (Claude Sonnet 4.5) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-romanian-nlu-rosent · How It Works · Data refreshed daily, snapshot 2026-10-20.