MathVision
multimodal
+
+
+
+
About
MathVision is a mathematical visual reasoning benchmark that evaluates AI models' ability to solve mathematical problems requiring visual understanding. This benchmark tests models' capacity to interpret mathematical diagrams, graphs, and geometric figures while performing mathematical reasoning, measuring the integration of visual perception and mathematical problem-solving skills in multimodal contexts.
+
+
+
+
Evaluation Stats
Total Models5
Organizations1
Verified Results0
Self-Reported5
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers
Score Distribution
5 models
Top Score
38.4%
Average Score
32.5%
High Performers (80%+)
0Top Organizations
#1Alibaba Cloud / Qwen Team
5 models
32.5%
+
+
+
+
Leaderboard
5 models ranked by performance on MathVision
License | Links | ||||
---|---|---|---|---|---|
Feb 28, 2025 | Apache 2.0 | 38.4% | |||
Jan 26, 2025 | tongyi-qianwen | 38.1% | |||
Dec 25, 2024 | Qwen | 35.9% | |||
Jan 26, 2025 | Apache 2.0 | 25.1% | |||
Mar 27, 2025 | Apache 2.0 | 25.0% |