OmniGen2 is an efficient multimodal generation model that combines visual language models and diffusion models, enabling functions such as visual understanding, image generation, and editing. Its open-source nature provides researchers and developers with a strong foundation to explore personalized and controllable AI generation.