MM-Thinker-72B-Preview

Model Overview

• MM-thinker-72B-Preview improves visual reasoning upon Qwen2.5-VL-72B-Instruct model.

• As of April 3rd, 2025, MM-thinker-72B-Preview achieves superior results on various visual reasoning benchmarks (MathVision,MathVista, MathVerse, MMMU-Pro, EMMA, MEGA).

Evaluation

We will release a code repository for VLM evaluation. It supports RL training with simple rule-based rewards, meanwhile aligning with LLM-Judge results.

Stay tuned!

Citation

If you find our model useful, please consider citing:

@misc {MM-Thinker-72B,
    author       = { Wang, Haozhe and Lin, Fangzhen and Chen, Wenhu },
    title        = { MM-Thinker-72B },
    year         = 2025,
    url          = { https://huggingface.co/TIGER-Lab/MM-Thinker-72B},
    publisher    = { Hugging Face }
}