InternVL 2.5 is a series of multimodal large language models launched by OpenGVLab, featuring significant enhancements in training strategies, testing strategies, and data quality improvements over InternVL 2.0. This series can process image, text, and video data, demonstrating capabilities in multimodal understanding and generation, positioning it at the forefront of the multimodal AI field. The InternVL 2.5 series provides robust support for multimodal tasks with its high performance and open-source attributes.