German LLM Index - MMLU-ProX (11,759 questions): leaderboard
German Language LLM Index (dach.peerbench.ai, a PeerBench project, 2026), which reruns every model itself with lm-evaluation-harness: chat template, temperature 0, up to 8,192 generated tokens, thinking models set to minimal reasoning effort where the API allows it and other models run as served. Labels carry (Reasoning off) when the site classed the run as reasoning-off (at most ten reasoning tokens, or under 0.1% of its completion tokens); other rows reasoned at the provider default or at the effort the label names. German MMLU-ProX: professionally translated MMLU-Pro questions with up to ten options across 14 subjects. This board holds runs on the full 11,759-question German set, before the v1.1 manifest removed 22 questions (duplicated answer keys from translation collisions, one ambiguous item, one collapsed notation). Accuracy (%), higher is better. Source: https://dach.peerbench.ai/
Metric: Accuracy (%). Source: dach.peerbench.ai. Saturation forecast: Around February 2027. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Argon | 91.61 | #9 |
| 2 | Gemini 3.1 Pro (Preview) | 89.18 | #64 |
| 3 | Gemini 3.8 Flash | 89.01 | #30 |
| 4 | Gemini 3.7 Flash | 88.21 | #24 |
| 5 | GPT-5.5 (Low) | 87.18 | #33 (GPT-5.5) |
| 6 | Gemini 3 Flash (Preview) (Non-reasoning) | 86.32 | #90 (Gemini 3 Flash (Preview)) |
| 7 | Gemini 3.1 Flash Lite (Non-reasoning) | 82.16 | #200 (Gemini 3.1 Flash Lite) |
| 8 | Gemma 4 31B (IT) (Non-reasoning) | 82.1 | #178 (Gemma 4 31B (IT)) |
| 9 | DeepSeek V4 Pro (Non-reasoning) | 80.76 | #112 (DeepSeek V4 Pro) |
| 10 | Qwen 3.6 35B A3B (Non-reasoning) | 80.05 | #238 (Qwen 3.6 35B A3B) |
| 11 | Gemini 2.5 Flash (Non-reasoning) | 79.62 | #255 (Gemini 2.5 Flash) |
| 12 | Gemma 4 26B A4B (IT) (Non-reasoning) | 78.19 | #293 (Gemma 4 26B A4B (IT)) |
| 13 | Claude Haiku 4.5 (Non-reasoning) | 75.26 | #287 (Claude Haiku 4.5) |
| 14 | DeepSeek V4 Flash (Non-reasoning) | 74.89 | #155 (DeepSeek V4 Flash) |
| 15 | Qwen 3.5 9B (Non-reasoning) | 73.37 | #391 (Qwen 3.5 9B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=german-llm-index-mmlu-prox-11-759-questions · How It Works · Data refreshed daily, snapshot 2026-11-05.