Image source: StabilityAI
Like the previous SD3 version, Stable Diffusion 3.5 is available in three configurations: Large (8B), Large Turbo (8B) and Medium (2.6B). All configurations are optimized to run on typical consumer hardware and can be customized. In its press release, StabilityAI admitted that the Stable Diffusion 3 Medium model, released in June, did not fully meet the standards and expectations of the community. “After listening to valuable feedback, rather than a quick fix, we decided to take the time to develop a version that advances our mission of transforming visual media,” the company said.
New models are focused on flexible customization, high performance and variety of results. Stylistic settings are supported, including photography and painting. You can also use hashtags to indicate a specific style, such as boho, impressionism or modern. You can also highlight keywords in the query to get more realistic images.
The Stable Diffusion 3.5 Large model leads the market in terms of compliance and image quality. The Turbo model has the minimum time to display results. Medium outperforms other models when it comes to balancing image quality and relevance, making it what the company claims is the most effective choice for content creation.
All three configurations are freely available under the Stability AI Community License. For commercial use, an Enterprise License is required.