Stable Diffusion 3.5
the six Ws · specification
Stability AI released Stable Diffusion 3.5.
Stable Diffusion 3.5 is a family of multimodal diffusion transformer text-to-image models, Large, Large Turbo, and Medium, offering improved prompt adherence and image quality over prior SD versions.
Hosted on Hugging Face under stabilityai/stable-diffusion-3.5 and on GitHub under Stability-AI.
Released in October 2024.
It restores fully open, commercially permissive weights for high-quality text-to-image generation after licensing controversy around SD3.
Built on the MMDiT architecture, runnable through Hugging Face diffusers or ComfyUI.
Released under the Stability AI Community License, free for individuals and businesses under 1M dollars annual revenue; weights gated behind license acceptance.
works with