CulturalMenuBench - Binary Verification: leaderboard
CulturalMenuBench (arXiv 2609.03526v1): culinary multimodal reasoning over 587 dishes from 18 regions with dish and step-by-step cooking images, ingredients and procedure text, prompted in each cuisine's language. Table 5 application probe: yes/no verification of dish-attribute pairings across 10 tasks for 12 models; accuracy percent, chance 50, higher is better.
Metric: Accuracy (%; yes/no pairing verification, chance 50). Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 78.6 | #105 |
| 2 | Gemini 2.5 Pro | 78.2 | #159 |
| 3 | GPT-5.1 (2025-11-13) | 78 | #107 |
| 4 | GPT-5 Mini (2025-08-07) | 76.1 | #175 |
| 5 | Claude Opus 4.6 | 72.5 | #67 |
| 6 | Claude Sonnet 4 | 72.4 | #207 |
| 7 | Qwen 3.6 Plus | 72.2 | #124 |
| 8 | Kimi K2.5 | 70.7 | #152 |
| 9 | Qwen 3.5 35B A3B | 69 | #268 |
| 10 | GPT-5 Nano | 68.9 | #439 |
| 11 | Gemma 4 26B A4B (IT) | 64.4 | #293 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=culturalmenubench-binary-verification · How It Works · Data refreshed daily, snapshot 2026-11-05.