Stable Diffusion 3 is Stability AI's third major text-to-image model family, built on a new transformer design that is much better at following prompts and spelling text.
In plain words
Older image models often garbled words inside pictures and mixed up prompts with several subjects. Stable Diffusion 3 uses a new transformer, the Multimodal Diffusion Transformer, that keeps separate weights for image and text tokens but lets information flow both ways. It is trained with rectified flow, a method that links noise and images along a straight path. The family was announced in sizes from 800 million to 8 billion parameters, and later updated as Stable Diffusion 3.5.
A quick example
Ask for "a café chalkboard that says 'soup of the day: tomato'". Earlier models often misspelled the sign; this design was built to render such text more accurately.
Where you'll meet it
- Text-to-image generation, including images with written text.
- Self-hosted image tools under Stability AI's licence.
Related
Sources
- PaperScaling Rectified Flow Transformers for High-Resolution Image Synthesis, Esser et al. (arXiv) · read 29 Sept 2026
- DocsStable Diffusion 3, Stability AI · read 29 Sept 2026
- Docsstabilityai/stable-diffusion-3.5-large, Stability AI (Hugging Face) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.