OpenVLM MathVista - VQA: leaderboard
OpenCompass OpenVLM MathVista visual question answering subcategory score.
Metric: Accuracy (%). Source: huggingface.co. Source: huggingface.co. Saturation forecast: Around March 2028. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 68.2 | #152 |
| 2 | GPT-5 | 65.9 | #98 |
| 3 | GPT-4o ChatGPT | 64.2 | |
| 4 | InternVL3-78B | 63.7 | #352 |
| 5 | InternVL3-38B | 62.6 | #405 |
| 6 | GPT-5 Mini (2025-08-07) | 62 | #172 |
| 7 | Doubao-1.5-Pro | 61.5 | #287 |
| 8 | Qwen 2 VL 72B | 59.8 | #460 |
| 9 | Gemini 2.0 Flash | 59.2 | #342 |
| 10 | InternVL2.5-8B-BoN-8 | 58.7 | #639 |
| 11 | InternVL2.5-78B-CoT | 58.7 | #469 |
| 12 | GPT-4.1 (2025-04-14) | 58.1 | #250 |
| 13 | VLM-R1-3B-Math-0305 | 57.5 | #743 |
| 14 | Qwen 2 VL 7B | 57 | #750 |
| 15 | InternVL2.5-78B | 57 | #431 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mathvista-vqa · How It Works · Data refreshed daily, snapshot 2026-10-20.