Stability unveils Stable Diffusion 3 Medium: advanced image generator with 2B parameters

Stability AI has launched its most advanced image generator, Stable Diffusion 3 Medium, which boasts two billion parameters for enhanced image quality. The initial Stable Diffusion 3 release was previewed back on Feb 22, with public availability via an API on April 17, and now this new model version excels in creating photorealistic images of hands and faces and generating accurate text without errors. It also effectively adheres to complex prompts and understands spatial relationships within images.

Despite its advanced capabilities, Stable Diffusion 3 Medium remains compact, making it suitable for individual computing systems and enterprise-tier GPUs. The model is particularly adept at extracting nuanced details from small datasets, offering high customization potential. Stability AI has open-sourced the model's weights under a free non-commercial license available through Hugging Face, with commercial licensing information provided directly by the company.

Users can access Stable Diffusion 3 Medium via Stability AI’s API, the Stable Assistant chatbot, and on Discord through Stable Artisan.