EuroEval Bulgarian NLU - Cinexio: leaderboard
The "Cinexio" subtask of EuroEval Bulgarian NLU. Bulgarian-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: Around January 2027. 212 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Bielik-11B-v2.3-Instruct | 61.38 | #693 |
| 2 | Qwen3.5-4B-Base | 61.35 | #798 |
| 3 | Qwen 3.5 27B FP8 | 61.2 | #164 |
| 4 | Qwen 3.5 35B A3B Base | 60.79 | #310 |
| 5 | Qwen 3.5 35B A3B FP8 | 60.43 | #362 |
| 6 | Ministral-3-8B-Instruct-2512 | 59.43 | #661 |
| 7 | Ministral-3-14B-Reasoning-2512 | 59.07 | #540 |
| 8 | Qwen 3.5 2B | 58.3 | #1148 |
| 9 | Qwen 3.5 9B | 57.78 | #379 |
| 10 | Gemma 4 26B A4B | 57.15 | #358 |
| 11 | Gemini 2.5 Flash Lite (Non-reasoning) | 57.04 | #422 (Gemini 2.5 Flash Lite) |
| 12 | GPT-4.1 | 56.99 | #248 |
| 13 | gemma-4-E4B-it | 56.94 | #566 |
| 14 | GPT-5.2 | 56.92 | #110 |
| 15 | Claude Sonnet 4.5 (Non-reasoning) | 56.6 | #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-bulgarian-nlu-cinexio · How It Works · Data refreshed daily, snapshot 2026-10-20.