math benchmark
MathVista
MathVista evaluates mathematical reasoning of foundation models in visual contexts. It consists of 6,141 examples derived from 28 existing multimodal datasets and 3 newly created datasets (IQTest, FunctionQA, and PaperQA), combining challenges from diverse mathematical and visual tasks to assess models' ability to understand complex figures and perform rigorous reasoning.
Updated Aug 11, 2026
Models39
Model coverage39
MetricScore
EvidenceB
MathVista Ranking
Higher score ranks better on this benchmark.
MathVista Score Distribution
A closer view of the leading scores on this benchmark.
MathVista Highlights
The leading models and scores on this benchmark.
What is MathVista?
What MathVista measures and how its scores work.
MathVista evaluates mathematical reasoning of foundation models in visual contexts. It consists of 6,141 examples derived from 28 existing multimodal datasets and 3 newly created datasets (IQTest, FunctionQA, and PaperQA), combining challenges from diverse mathematical and visual tasks to assess models' ability to understand complex figures and perform rigorous reasoning.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- MathVista
- Modality
- multimodal
- Primary category
- math
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- mathvista|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about MathVista.
Which model scores highest on MathVista?
Seed 2.1 Pro is currently ranked first with 90.7%.
What does MathVista measure?
MathVista evaluates mathematical reasoning of foundation models in visual contexts. It consists of 6,141 examples derived from 28 existing multimodal datasets and 3 newly created datasets (IQTest, FunctionQA, and PaperQA), combining challenges from diverse mathematical and visual tasks to assess models' ability to understand complex figures and perform rigorous reasoning.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
39 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.