Grip on LLMs - Dutch tinyMMLU: leaderboard

Metric: Accuracy (%; GP-IRT estimate from 100 tinyBenchmarks items, Dutch machine translation). Source: arxiv.org. 31 models tracked.

Top models

#ModelScore
1GPT-4o70
2Mistral Large 369
3GPT-568
4Mistral Medium 366
5GPT-5 Mini65
6Gemma 3 27B (IT)60
7Qwen 3 32B60
8DeepSeek R1 Distill Qwen 32B57
9GPT-4o Mini56
10Gemma 3 12B (IT)54
11GPT-5 Nano54
12aya-expanse-32B54
13Mistral Small 352
14Qwen 3 8B50
15GPT-OSS-20B47

Interactive version: theaggregate.ai/benchmark?slug=grip-on-llms-dutch-tinymmlu · How It Works · Data refreshed daily, snapshot 2026-09-19.