Stability AI ships Stable Diffusion 3 Medium weights
The 2-billion-parameter weights ran on consumer GPUs but were licensed non-commercially, with a separate paid Enterprise tier for large-scale business use.
- Models & capabilities
- Open weights & ecosystem
- Minor
Stability AI released the weights for Stable Diffusion 3 Medium, a 2-billion-parameter text-to-image model, following a preview announcement of the SD3 architecture in February 2024. The company said the model produced sharper detail and lighting than its predecessors and, notably, handled text rendering within images — spelling, kerning and letter spacing — more reliably than earlier Stable Diffusion versions, an area where diffusion models had generally struggled.
The model’s modest size let it run on consumer hardware; Stability said Nvidia RTX GPU optimisation roughly doubled inference speed, and it published equivalent AMD optimisations. Weights were published on Hugging Face under a “Stability AI Community License,” permitting free use for research and by small organisations, while businesses above a revenue threshold were directed to a paid Enterprise licence — a two-tier structure Stability had adopted for its recent releases rather than the fully open licences of its earlier models.
The release continued a pattern in image generation of small, efficient open-weight models competing against much larger closed systems from OpenAI (DALL-E 3) and Midjourney, with accessibility and licensing terms as much a point of competition as raw output quality.