Stable Diffusion 3 is a text-to-image model from Stability AI. It is a diffusion transformer, which is a new type of architecture similar to the one used in the OpenAI Sora model.