This spring, Stability AI introduced stable 3 diffusionAnd now, half year later, the next model, Stable 3.5has been published with notable improvements and new features for the art generated by AI. This version includes three model variants: Stable 3.5 large, large turbo and medium broadcast – all available under the AI Community license. These models are customizable, optimized for consumer equipment and open for non -commercial and limited commercial use.
Stable Diffusion 3.5 provides significant progress in terms of speed, performance and accessibility. According to Stability IA, this version reflects their commitment to democratize the generation of high quality images and focused on AI. One of the most impactful improvements of this version is the quality and resolution of the improved image, fueled by advanced algorithms that produce sharper and more detailed images, which facilitates the creation of high quality realistic works of art.
What to expect:
- Stable 3.5 large broadcast: With 8.1 billion parameters, this model offers a very precise generation of images ideal for professional use to resolutions up to 1 megapixel.
- Stable 3.5 Grand Turbo broadcast: A rationalized version of the large model, offering a quality similar to accelerated speeds in four -step prompts.
- Stable 3.5 medium broadcast: Optimized for consumer equipment, this model provides a high -quality generation of images on standard GPUs, requiring as little as 9.9 GB of VRAM.
Stable Diffusion 3.5 also highlights AI stability on flexible and various applications. By incorporating the standardization of the request keys in the transformer blocks, AI stability allows more fluid training and customizable adaptations. In addition, these new models are designed to produce various outings with a minimum of incentive, to adapt to a wide range of styles of user and use cases.
Released under the license of the AI of Stability AI, Stable Diffusion 3.5 is available for free for non -commercial use, including research. In addition to self-hosting via FaceStability AI has rendered the model accessible via several platforms, including the AI stability API, the reply, Deepinfra and the Comfyui. Committed to promoting responsible use of AI, stability AI has introduced strict safety protocols to avoid improper use.
In the near future, AI stability plans to publish control links to provide advanced control options for professional users, improving the range of potential applications.
