MathVista
Active
6,141 multimodal math problems combining diagrams, charts, and figures - the canonical "math + vision" benchmark.
- Publisher
- University of California, Berkeley
- Capabilities
- MathImage Understanding
- Domain
- multimodal
- Format
- HF Dataset
- Size
- 6141 tasks
- License
- CC-BY-SA-4.0
- Published
- May 2026
- Notable for
- Benchmark for evaluating math and image understanding in the multimodal domain.
- Canonical
- mathvista.github.io
Cite
Notes
Only stored in your browser.
Related tools
2Implementations, trainers, datasets and scaffolds linked to this eval.
FAQ
- What is MathVista?
- 6,141 multimodal math problems combining diagrams, charts, and figures - the canonical "math + vision" benchmark.
- What capabilities does MathVista test?
- MathVista evaluates math, image understanding.
- How can a model improve its MathVista score?
- Tools linked to MathVista on Sophon include Multimodal RL Env (Community), NuminaMath - RL environments, datasets, and scaffolds that target this eval.
- What license is MathVista under?
- MathVista is available under CC-BY-SA-4.0.