MathVision

A visual mathematics benchmark that tests whether a model can solve math problems grounded in diagrams, equations, figures, and other visual inputs.

In index
Multimodal
Best in multimodal

Models24

Top result95.2%Qwen3.8 Max

Top-3 spread2.4 pts

Year—

Result and price

16 of 24 results

Results

24 results

#ModelResult

01Qwen3.8 Max95.2%

02Kimi K394.3%

03Inkling 2 Small92.8%

04Seed 2.1 Pro92.6%

05Qwen3.8-Omni-Flash91.8%

06Qwen3.8-Flash-Next90.6%

07Qwen3.7 Plus90.3%

08Seed 2.1 Turbo90.1%

09Qwen3.8-27B90.0%

10Qwen3.5 397B88.6%

11Qwen3.6 Plus88.0%

12dots3-note Preview87.7%

13Kimi K2.687.4%

14Gemini 3 Pro86.6%

15Qwen3.5-122B-A10B86.2%

16Qwen3.5-27B86.0%

17Kimi K2.584.2%

18Qwen3.5-35B-A3B83.9%

19GPT-5.283.0%

20Gemma 4 12B79.7%

21Nemotron 3.5 Super VL 120B A12B78.2%

22Nemotron 3.5 Super VL 120B A12B NVFP477.8%

23Claude Opus 4.574.3%

24LFM3-VL-2B51.6%