SommBench - Wine Theory QA (German): leaderboard
Metric: Wine theory QA accuracy (%) with German prompts: accuracy on SommBench's 128 four-option wine theory questions (WSET-style oenology), answered with a single letter, zero-shot, temperature 0, one run; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 97 | #145 |
| 2 | GPT-5 | 97 | #91 |
| 3 | Grok 4 | 97 | #169 |
| 4 | Grok 4 Fast | 94 | #242 |
| 5 | Gemini 2.5 Flash | 93 | #237 |
| 6 | GPT-4o | 91 | #333 |
| 7 | GPT-4.1 | 91 | #240 |
| 8 | GPT-OSS-120B (High) | 86 | #330 (GPT-OSS-120B) |
| 9 | GPT-4.1 Mini | 85 | #346 |
| 10 | GPT-OSS-120B (Medium) | 84 | #330 (GPT-OSS-120B) |
| 11 | GPT-4o Mini | 83 | #588 |
| 12 | Gemini 2.5 Flash Lite | 83 | #413 |
| 13 | GPT-OSS-120B (Low) | 83 | #330 (GPT-OSS-120B) |
| 14 | GPT-4.1 Nano | 72 | #716 |
| 15 | GPT-OSS-20B (Medium) | 70 | #499 (GPT-OSS-20B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=sommbench-wine-theory-qa-german · How It Works · Data refreshed daily, snapshot 2026-10-11.