OpenVLM MathVista - FQA: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 284 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 82.2 |
| 2 | GPT-5 | 81.8 |
| 3 | GPT-5 Mini (2025-08-07) | 79.9 |
| 4 | GPT-4.1 Mini | 79.2 |
| 5 | InternVL3-78B | 78.1 |
| 6 | QVQ-72B-Preview | 77.7 |
| 7 | Taichu-VLR-7B | 77 |
| 8 | Qwen 2 VL 72B | 76.6 |
| 9 | InternVL3-38B | 75.5 |
| 10 | Doubao-1.5-Pro | 75.5 |
| 11 | VLAA-Thinker-Qwen2.5VL-7B | 75.5 |
| 12 | GPT-4o ChatGPT | 74.3 |
| 13 | GPT-4.1 (2025-04-14) | 73.6 |
| 14 | Taichu-VLR-3B | 73.6 |
| 15 | Claude 3.5 Sonnet (20241022) | 72.5 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mathvista-fqa · How It Works · Data refreshed daily, snapshot 2026-09-05.