DiningBench - Fine-Grained Classification: leaderboard
Metric: Accuracy (%) choosing the dish among eight same-menu candidates (seven hard distractors) from multi-view user photos, over the 2,884 DiningBench classification samples; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 81.83 |
| 2 | Gemini 3 Pro (Preview) | 81.55 |
| 3 | Gemini 2.5 Pro | 73.51 |
| 4 | Gemini 2.5 Flash | 71.01 |
| 5 | GPT-5 | 70.18 |
| 6 | GPT-4.1 | 68.59 |
| 7 | Qwen 3 VL 30B A3B Instruct | 65.43 |
| 8 | Qwen 2.5 VL 72B Instruct | 65.29 |
| 9 | GPT-4o | 65.26 |
| 10 | O4 Mini | 64.81 |
| 11 | Qwen 3 VL 8B Instruct | 64.15 |
| 12 | Qwen 3 VL 30B A3B (Thinking) | 61.34 |
| 13 | Qwen 2.5 VL 32B Instruct | 61.17 |
| 14 | Qwen 2.5 VL 7B Instruct | 60.85 |
| 15 | Qwen 3 VL 4B Instruct | 60.06 |
Interactive version: theaggregate.ai/benchmark?slug=diningbench-fine-grained-classification · How It Works · Data refreshed daily, snapshot 2026-10-07.