MEGA-Bench Task - Bongard Problem: leaderboard
Metric: Task Score (%). Source: huggingface.co. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL2.5-2B | 81.6 |
| 2 | Pixtral-12B | 34.2 |
| 3 | Claude 3.5 Sonnet (20240620) | 31.6 |
| 4 | Gemini 1.5 Flash (002) | 28.9 |
| 5 | InternVL2.5-78B | 24.6 |
| 6 | Claude 3.5 Sonnet (20241022) | 23.7 |
| 7 | GPT-4o | 21.1 |
| 8 | Qwen 2 VL 72B | 21.1 |
| 9 | Qwen 2 VL 7B | 21.1 |
| 10 | GPT-4o Mini | 18 |
| 11 | Gemini 1.5 Pro (002) | 15.8 |
| 12 | InternVL2-8B | 13.2 |
| 13 | Qwen 2 VL 2B | 10.5 |
| 14 | Aquila-VL-2B | 10.5 |
Interactive version: theaggregate.ai/benchmark?slug=mega-bench-task-bongard-problem · How It Works · Data refreshed daily, snapshot 2026-09-05.