Stable Diffusion XL-base-1

A 3.5 billion parameter latent diffusion model engineered for high-resolution 1024x1024 image synthesis and complex spatial reasoning.

Stable Diffusion XL (SDXL) 1.0 represents a significant architectural leap over its predecessors, utilizing a base model with a 3x larger UNet backbone to handle native 1024x1024 resolutions. It employs a dual-conditioning strategy (OpenCLIP-ViT/G and CLIP-ViT/L) to improve prompt adherence and visual fidelity. This specific base-1.0 iteration excels at rendering legible text, intricate human anatomy, and photorealistic lighting without requiring extensive prompt engineering. It serves as the definitive open-weights foundation for fine-tuning professional-grade creative workflows.

https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0