We-Math
A multimodal math benchmark for visually grounded mathematical reasoning and answer generation.
5 models
Higher is better.
- 1Qwen3.5 397BAlibaba87.9
- 2Gemini 3 ProGoogle86.9
- 3Kimi K2.5Moonshot AI84.7
- 4GPT-5.2OpenAI79.0
- 5Claude Opus 4.5Anthropic70.0
About this benchmark
- Metric
- Multimodal mathematical reasoning
- Tasks
- Visually grounded math problems
- Authors
- Qwen (2026)
Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.