EuroEval Finnish NLU - Scandisent FI: leaderboard
The "Scandisent FI" subtask of EuroEval Finnish NLU. Finnish-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. 377 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | gemma-3-27B-pt | 93.78 | #502 |
| 2 | Llama 2 70B Base | 93.51 | #1025 |
| 3 | Llama 3.1 405B | 93.31 | #466 |
| 4 | Gemma 3n E4B | 93.19 | #878 |
| 5 | Gemma 2 27B | 93.12 | #806 |
| 6 | gemma-3-12B-pt | 92.87 | #686 |
| 7 | Apertus-70B-Instruct-2509 | 92.75 | #820 |
| 8 | Llama 3.1 Nemotron 70B Instruct | 92.66 | #554 |
| 9 | Llama 3.1 70B | 92.64 | #592 |
| 10 | gemma-3-4B-pt Base | 92.64 | #1134 |
| 11 | Llama 3.1 70B Instruct | 92.62 | #561 |
| 12 | Llama 3.1 405B Instruct FP8 | 92.59 | #389 |
| 13 | Llama 4 Scout Instruct | 92.52 | #539 |
| 14 | Qwen 2.5 72B Instruct | 92.47 | #454 |
| 15 | Claude 3.5 Haiku (20241022) | 92.47 | #588 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-finnish-nlu-scandisent-fi · How It Works · Data refreshed daily, snapshot 2026-10-20.