🇩🇪 Best AI for German
Use this one
- Best overall: gpt-6-astra. Statistically close: gpt-6.1-sol.
- Best open-weight (self-host / cheap): zai-org/GLM-5.3-Flash (rank score 1.42).
- Best for writing (summaries + grammar, top 40 models): zai-org/GLM-5.3-Flash (54.83).
- Translation & agent tasks: no per-model German benchmark data yet in EuroEval's public generative board, so we don't rank them. See gaps.
| # | Model | Rank score ↓ | Writing ↑ | Understanding ↑ |
|---|---|---|---|---|
| 1 | gpt-6-astra | 1.35 | 53.59 | 73.08 |
| 2 | gpt-6.1-sol | 1.39 | 53.12 | 71.69 |
| 3 | zai-org/GLM-5.3-Flash open | 1.42 | 54.83 | 64.41 |
| 4 | gemini/gemini-3.7-flash | 1.44 | 51.04 | 73.81 |
| 5 | gpt-6-sol | 1.52 | 49.78 | 66.38 |
| 6 | gemini/gemini-3.6-flash | 1.52 | 48.45 | 73.0 |
| 7 | gpt-5.6-sol | 1.58 | 48.64 | 69.55 |
| 8 | deepseek/deepseek-flash#no-thinking | 1.62 | 52.38 | 58.39 |
| 9 | deepseek/deepseek-flash#low | 1.64 | 51.64 | 61.11 |
| 10 | Qwen/Qwen3.8-27B open | 1.71 | 51.24 | 54.19 |
| 11 | gpt-5.6-terra | 1.72 | 47.72 | 62.11 |
| 12 | mistralai/Mistral-Small-3.1-24B-Base-2503 open | 1.73 | 52.95 | 48.51 |
| 13 | Qwen/Qwen3.6-27B-FP8 open | 1.73 | 47.09 | 54.69 |
| 14 | zai-org/GLM-5.3-Flash open | 1.73 | 50.31 | 61.72 |
| 15 | gemini/gemini-3-flash-preview#thinking | 1.74 | 49.47 | 71.77 |
Source: EuroEval German generative leaderboard, retrieved 2026-10-15. 188 models evaluated. Datasets: germanquad, germeval, include_de, mlsum_de, multiloko_de, sb10k, scala_de, winogrande_de.
The smartest model isn't always deployable.
Get ahead of the constraints with The Signal.