FLUX.1-dev

FLUX.1-dev is a 12 billion parameter rectified flow transformer designed to generate images from text prompts. Built by Black Forest Labs, it sits just below the flagship FLUX.1 pro in their model hierarchy, delivering cutting-edge image quality that rivals closed-source alternatives while keeping its architecture open for inspection. The model emphasizes strong prompt-following capabilities, meaning it interprets and executes complex text instructions with precision. Beyond standalone image generation, FLUX.1-dev serves as the foundation for an expanding ecosystem of specialized tools—FLUX.1 Fill for inpainting and outpainting, FLUX.1 Depth for structure-guided generation, FLUX.1 Canny for edge-guided creation, and FLUX.1 Redux for blending input images with text prompts.

What sets FLUX.1-dev apart is its training approach: guidance distillation, a technique that compresses the model's decision-making process to be more efficient without sacrificing output quality. This makes it practical for real-world deployment while remaining accessible for research experimentation. The model ships with a reference implementation and sampling code on GitHub, inviting the community to explore, study, and build innovative workflows around it. Though the base license carries non-commercial terms, outputs generated by the model can be used for personal, scientific, and commercial purposes—bridging the gap between open research and real-world applications. FLUX.1-dev represents a deliberate balance: giving the research community a powerful, inspectable model while laying groundwork for artists and developers to create next-generation generative tools.