Translated data: An open-source project, MM-Interleaved, has achieved a new breakthrough in the field of multimodal large-scale generative models, garnering widespread attention from scholars. The project introduces an innovative feature synchronizer, which has set new state-of-the-art records in multiple tasks and expanded the application areas of various text-image and image generation tasks. The model performs exceptionally well during the pretraining phase and maintains a leading position even after fine-tuning for specific tasks, providing crucial support for the development of multimodal large models. It has taken a significant step towards comprehensive end-to-end unified modeling and training.