Stability AI The company published a blog post yesterday (October 22) announcing the launch of the Stable Diffusion 3.5, which marks theOpen Source AI Wenshengtu Modelsignificant progress.
Stable Diffusion 3.5 is available in Medium (released October 29th), Large and Large Turbo sizes, designed to meet the needs of scientific researchers, hobbyists, startups and enterprises, as described below:
- Stable Diffusion 3.5 Large:With a total of 8 billion parameters, superior quality and fast response, it is the most powerful model in the Stable Diffusion family and is ideally suited for professional application scenarios at 1 megapixel resolution.
- Stable Diffusion 3.5 Large Turbo:A lite version of Large, it generates high-quality images in just 4 steps and is significantly faster than Large, while following the cue word well.
- Stable Diffusion 3.5 Medium (to be released on October 29)With 2.5 billion parameters, an improved MMDiT-X architecture and training methodology, the model is designed to work "out-of-the-box" on consumer-grade hardware, balancing quality with ease of customization, and is capable of generating photographs with resolutions ranging from 250,000 to 2 megapixels.
Stability AI has indicated that this release did not fully meet our standards or our community's expectations, and will continue to polish and develop a more robust solution.
Stability AI has implemented a fairly liberal community license for this release; the models are available for non-commercial use at no charge and are open to businesses with annual revenues of less than $1 million, above which a separate licensing arrangement must be obtained.