SommBench: leaderboard
Metric: SommBench score (x 100, from -33.3 to 100): mean of the wine theory QA accuracy and the wine feature completion accuracy (each averaged over eight languages) and the food and wine pairing MCC, zero-shot, temperature 0, one run; higher is better. Source: arxiv.org. Saturation forecast: Around April 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 65 | #237 |
| 2 | GPT-4.1 | 59 | #240 |
| 3 | GPT-4o | 57 | #333 |
| 4 | Gemini 2.5 Pro | 57 | #145 |
| 5 | GPT-5 | 57 | #91 |
| 6 | GPT-4.1 Mini | 54 | #346 |
| 7 | Grok 4 | 53 | #169 |
| 8 | Grok 4 Fast | 52 | #242 |
| 9 | GPT-4o Mini | 51 | #588 |
| 10 | Gemini 2.5 Flash Lite | 49 | #413 |
| 11 | GPT-OSS-120B (Low) | 47 | #330 (GPT-OSS-120B) |
| 12 | GPT-OSS-120B (Medium) | 47 | #330 (GPT-OSS-120B) |
| 13 | GPT-OSS-20B (Low) | 42 | #499 (GPT-OSS-20B) |
| 14 | GPT-OSS-20B (Medium) | 42 | #499 (GPT-OSS-20B) |
| 15 | GPT-4.1 Nano | 41 | #716 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=sommbench · How It Works · Data refreshed daily, snapshot 2026-10-11.