EuroEval German NLU - GermEval: leaderboard
The "GermEval" subtask of EuroEval German NLU. German-language EuroEval natural-language-understanding suite, separating NLU task performance from the broader all-task EuroEval aggregate.
Metric: Micro F1 excluding MISC (%). Source: euroeval.com. Saturation forecast: Around January 2027. 323 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 235B A22B 2507 FP8 (Thinking) | 78.33 | #246 (Qwen 3 235B A22B 2507 FP8) |
| 2 | Qwen 3 4B 2507 (Thinking) | 78.27 | #538 (Qwen 3 4B 2507) |
| 3 | GPT-5.4 Mini (2026-03-17) (High) | 77.95 | #169 (GPT-5.4 Mini (2026-03-17)) |
| 4 | Claude Sonnet 4.5 (Thinking) | 77.06 | #142 (Claude Sonnet 4.5) |
| 5 | NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 | 75.4 | #679 |
| 6 | GPT-5 | 75.28 | #98 |
| 7 | Gemma 4 31B (IT) | 75.25 | #176 |
| 8 | OLMo 3.1 32B (Thinking) | 74.37 | #632 (OLMo 3.1 32B) |
| 9 | Gemma 4 26B A4B (IT) | 74.21 | #297 |
| 10 | Claude Sonnet 4.5 (Non-reasoning) | 74.2 | #142 (Claude Sonnet 4.5) |
| 11 | Qwen 3 Next 80B A3B (Thinking) | 74.11 | #320 (Qwen 3 Next 80B A3B) |
| 12 | GPT-5 (2025-08-07) (High) | 73.8 | |
| 13 | Qwen 3 14B | 73.49 | #537 |
| 14 | Qwen 3 30B A3B | 73.17 | #493 |
| 15 | Llama 3.1 405B Instruct FP8 | 72.32 | #389 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-german-nlu-germeval · How It Works · Data refreshed daily, snapshot 2026-10-20.