Stable Diffusion XL 0.9: dual-encoder and 1024x1024 resolution

In one sentence Stability AI launches SDXL 0.9 beta with dual-encoder architecture and separate refiner model for photographic-quality 1024x1024 images.

Verified Official source

ShareLinkedIn X

Stable Diffusion XL is the heavily upgraded version of the model everyone knows. The main novelty is resolution: instead of 512x512 pixels like the original model, SDXL generates 1024x1024 images — double in each dimension.

Stability AI also introduced two models working together: a "base" that generates the image structure, and a "refiner" that adds detail and quality in the final stage. The visual result is significantly better, with more readable text in images and more correct anatomy.

Version 0.9 is a beta, but the quality jump over SD 1.5 is visible to the naked eye. For the first time, an open source model visually competes with Midjourney.