SDXL increases the U-Net parameter count to 3.5 billion, adopts dual text encoders (OpenCLIP ViT-bigG and CLIP ViT-L), and uses a two-stage generation architecture (base + refiner).