InternVL 2.5 is an advanced series of multimodal large language models. Building on InternVL 2.0, it enhances training and testing strategies and improves data quality while maintaining its core architecture. This model integrates the newly pre-trained InternViT with various large language models, such as InternLM 2.5 and Qwen 2.5, utilizing a randomly initialized MLP projector. InternVL 2.5 supports multiple images and video data, employing dynamic high-resolution training methods to provide better performance when processing multimodal data.