Stable Diffusion 3 Medium
Stable Diffusion 3 Medium is an earlier MMDiT image model whose weights remain published. Its files include several packages with the same diffusion model but different bundled text encoders.
Model design and access
The card identifies CLIP-L, OpenCLIP-G and T5-XXL text encoders. Removing T5 changes model behavior, even when it reduces resource needs. The ComfyUI and Diffusers packages are forms of the same release rather than distinct image models.Stable Diffusion 3 Medium official model documentation
The SD 3.0 API was deprecated on 17 April 2025. Requests for sd3-medium are redirected to sd3.5-medium. The published 3 Medium weights remain a way to run the earlier model itself.Stability AI API release notes
Practical use
Practical guidance. Keep these weights when an existing local workflow depends on their response or adapters. An old API name no longer establishes that the old model ran. If output consistency matters, verify the installed checkpoint and encoder package, rather than relying only on a familiar model label.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | sd3-mediumstable-diffusion-3-medium-diffusers |
| License | Stability AI Community License.Stable Diffusion 3 Medium official model documentation |
| Packaging | Full encoders, reduced-precision T5 or packages without T5.Stable Diffusion 3 Medium official model documentation |
Related models
- Stable Diffusion 3.5 Medium
Stability AI · Stable Diffusion
Stable Diffusion 3.5 Medium uses the MMDiT-X architecture for image generation.
Launch date not confirmedAPI · Downloadable