Stable Diffusion XL (SDXL) 1.0 Released: One of the Largest Parameter Counts Among Open Models
Stability AI released โSDXL 1.0,โ the latest flagship model in the Stable Diffusion lineup. Announced at AWS Summit New York, the model boasts one of the largest parameter counts among openly accessible image generation models, with major improvements to color, contrast, and shadow rendering.
Details
- Model scale: The base model has 3.5 billion parameters, while the base-plus-refiner ensemble pipeline totals 6.6 billion
- Resolution: Supports native 1024x1024 high-resolution output
- Image quality improvements: Vivid, accurate color rendering, along with significantly improved reproduction of contrast, lighting, and shadow compared to the previous generation
- Availability: Model weights and source code were published on Stability AIโs GitHub, alongside availability as an API through the Stability AI Platform
- License: Released under the CreativeML OpenRAIL++-M license
What happened next
SDXL significantly raised the performance bar for open-weight models and was widely adopted as the base for many fine-tuned derivative models and tools. This success raised expectations for the Stable Diffusion 3 series the following year, even as it coincided with a period when Stability AIโs business-side challenges began to surface.