InternVL2.5-MPO is an advanced series of large multimodal language models built on InternVL2.5 and Mixed Preference Optimization. This model integrates the newly added InternViT with various pretrained large language models, including InternLM 2.5 and Qwen 2.5, utilizing a randomly initialized MLP projector. The model architecture of InternVL2.5-MPO retains the same structure as InternVL 2.5 and its predecessors, following the 'ViT-MLP-LLM' paradigm. It supports multiple image and video data types and enhances model performance through mixed preference optimization (MPO), excelling in multimodal tasks.