SD 3 Medium

image

by Stability AI · Updated August 1, 2025

Generate with SD 3 Medium

Stable Diffusion 3 Medium uses a new MMDiT architecture for improved image quality, text rendering, and compositional understanding. A significant step forward in open-source image generation.

Best For

Text renderingImproved compositionNext-gen qualityOpen-source

Prompting Tips

  1. 1SD3 handles text in images significantly better than SDXL
  2. 2Uses a new prompt format — more natural language friendly
  3. 3CFG scale around 4-7 works well
  4. 4Supports both positive and negative prompts

Syntax & Constraints

Natural language prompts with optional weighting. Negative prompt supported. New architecture allows better understanding of spatial relationships.

Build Prompts for SD 3 Medium

Other Stable Diffusion Models

Related Guides