MathVista
Multimodal · 2023-10-03
MathVista evaluates mathematical reasoning with visual context across diverse task types and domains. Scores are reported as answer accuracy.
Top models (higher is better)
| Model | Score |
|---|---|
| o3 | 86.8 |
| O4 Mini | 84.3 |
| Llama 4 Maverick Instruct | 73.7 |
| GPT-4.1 Mini | 73.1 |
| Gemini 2.0 Flash | 73.1 |
| GPT-4.5 | 72.3 |
| GPT-4.1 | 72.2 |
| o1 (2024-12-17) | 71.8 |
| Llama 4 Scout Instruct | 70.7 |
| Mistral Small 3.1 24B Instruct 2503 | 68.9 |
| Gemma 3 27B IT | 67.6 |
| GPT-4o (2024-11-20) | 61.4 |
| GPT-4o Mini | 56.5 |
| GPT-4.1 Nano | 56.2 |