FLUX 3

FLUX 3 is a new multimodal foundation model developed by Black Forest Labs, designed to jointly learn from images, videos, and audio within a unified architecture. Its core mission is to build a comprehensive representation of the world, understanding how objects hold together, how things move, and how events sound, recognizing that individual modalities offer incomplete descriptions of reality. Built upon Black Forest Labs' proprietary Self-Flow approach, FLUX 3 efficiently aligns multimodal generation and understanding. This enables it to generate images and video with native audio jointly, from both pure text prompts and various input references such as images and existing videos. The model also extends its capabilities to action prediction, either natively integrated or serving as a dynamics-aware foundation for specialized action models. Currently available in Early Access, FLUX 3 represents a significant step towards developing real-world visual intelligence. Early results indicate strong performance in content creation and physical AI applications, with continuous improvements expected during its early access phase.