DiningBench - VQA: leaderboard
Metric: Accuracy (%) on the 804 DiningBench visual questions (cuisine technique, dietary advice, counterfactual and multi-image reasoning), answers judged by an LLM against the reference; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 90.42 |
| 2 | Gemini 2.5 Pro | 89.93 |
| 3 | Gemini 3 Flash (Preview) | 88.56 |
| 4 | GPT-5 | 86.94 |
| 5 | GPT-4.1 | 84.33 |
| 6 | Claude Sonnet 4.5 | 83.58 |
| 7 | GPT-4o | 80.6 |
| 8 | Qwen 3 VL 30B A3B Instruct | 80.6 |
| 9 | O4 Mini | 80.35 |
| 10 | Gemini 2.5 Flash | 78.36 |
| 11 | Qwen 2.5 VL 72B Instruct | 76.62 |
| 12 | Qwen 3 VL 30B A3B (Thinking) | 74.13 |
| 13 | Qwen 3 VL 8B Instruct | 72.76 |
| 14 | GPT-4o Mini | 71.39 |
| 15 | Qwen 2.5 VL 32B Instruct | 71.14 |
Interactive version: theaggregate.ai/benchmark?slug=diningbench-vqa · How It Works · Data refreshed daily, snapshot 2026-10-07.