BenGER: leaderboard
German-law benchmark for subsumption-based legal reasoning, evaluating model answers across Benchathon, ZJS, and doctrinal-principles corpora.
Metric: Aggregate Accuracy (%). Source: arxiv.org. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 77 | #58 |
| 2 | Claude Opus 4.7 | 75 | #49 |
| 3 | Claude Sonnet 4.6 | 74 | #90 |
| 4 | GPT-5.4 | 74 | #79 |
| 5 | Qwen 3.5 122B A10B | 73 | #173 |
| 6 | Gemini 3.1 Flash Lite | 71 | #191 |
| 7 | Qwen 3 235B A22B | 70 | #313 |
| 8 | Qwen 3.6 35B A3B | 68 | #226 |
| 9 | DeepSeek V4 Pro | 67 | #102 |
| 10 | GPT-5.4 Mini | 66 | #212 |
| 11 | DeepSeek V4 Flash | 64 | #137 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=benger · How It Works · Data refreshed daily, snapshot 2026-10-20.