Stable Diffusion 3.5 is an open image generation model family offering high prompt adherence, diverse styles, and flexible consumer hardware deployment.
What it is
Stable Diffusion 3.5 is an open-weights image generation model family created to provide high-quality visual outputs with strong prompt adherence. Available in multiple variants including Large, Large Turbo, and Medium, the models cater to researchers, hobbyists, startups, and enterprises looking to generate diverse visual media ranging from photography to 3D and line art.
Key features
- Multiple model variants including Large, Large Turbo, and Medium tailored for different performance and speed requirements
- Runs efficiently on standard consumer hardware, requiring 9.9 GB of VRAM for the Medium model
- Integrated Query-Key Normalization to stabilize training and simplify fine-tuning and LoRA creation
- Generates diverse human representations and wide-ranging artistic styles like 3D, painting, and photography
- Available via self-hosting through Hugging Face, GitHub, and multiple API deployment partners
Who it's for
Stable Diffusion 3.5 is built for developers, artists, researchers, startups, and enterprises wanting to generate or fine-tune custom image models on local hardware or cloud infrastructure.
Pricing
Free for non-commercial use and commercial use for organizations with up to $1M in annual revenue under the Stability AI Community License. Enterprise licenses are available for organizations exceeding $1M in annual revenue.
