OpenVLM AI2D - Parts of a Whole — leaderboard
Metric: Accuracy (%). Source: huggingface.co. 284 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 88.1 |
| 2 | GPT-5 (2025-08-07) | 86.2 |
| 3 | InternVL3-38B | 85.8 |
| 4 | InternVL2.5-78B | 85.8 |
| 5 | InternVL3-78B | 85 |
| 6 | GPT-4.5 | 84.6 |
| 7 | InternVL3-14B | 84.2 |
| 8 | GPT-5 Mini (2025-08-07) | 83.8 |
| 9 | Gemini 2.0 Flash | 82.8 |
| 10 | GPT-4.1 (2025-04-14) | 82.5 |
| 11 | GPT-4o ChatGPT | 82.5 |
| 12 | Qwen 2 VL 72B | 81.9 |
| 13 | Ovis2-8B | 81.7 |
| 14 | InternVL3-8B | 81.5 |
| 15 | GPT-4.1 Mini | 79.1 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-ai2d-parts-of-a-whole · How the rankings work · Data refreshed daily, snapshot 2026-07-22.