In the end of January 2026 Alibaba's Tongyi-MAI lab dropped its full version of Z-Image. The base model adds more control and range. It supports many art styles, photo styles, poses, and moods. You can reuse the same prompt and just change the seed to get very different results. That was hard to do with the Turbo version.
Negative prompts also work better here. You can remove traits like sadness or adjust how characters look. This gives you more control over outputs compared to Turbo.
This release is not the raw base model. The real base is called Zimage Omni Base. Z-Image-Omni-Base has not been released yet. It is described as a planned foundation model for both image generation and editing, meant to be a raw starting point for community fine-tuning once available. Z-Image which was released here - sits above Omni Base.
It can handle both image creation and editing. Tongyi Lab says it stays undistilled, which makes it better for training and LoRA work.
If you'd like to access this model, you can explore the following possibilities:
Weights