The Tongyi Lab from Alibaba introduced Z-Image-Turbo — a new open image generation model that operates without restrictions and creates realistic images even on graphics cards with 6 GB of memory. Users can run it locally even on laptops with RTX 2060 and get images in half a minute. The model has no censorship, so it can generate any content, including portraits of famous people and fictional characters.
Z-Image-Turbo has 6 billion parameters and uses the S3-DiT architecture, which allows combining textual and visual data for maximum efficiency. It accurately reproduces complex scenes, details skin, hair, and body proportions, and demonstrates a high level of understanding of style and spatial relationships. The model follows user instructions and makes almost no mistakes in complex compositions.
A separate advantage is the quality of text generation on images in English and Chinese. Z-Image-Turbo confidently handles inscriptions, which was previously a problem for other open models. Users note that prompts in Chinese yield even better results, and the “prompt enhancer” feature helps achieve the desired image.
The model quickly gained popularity among the community: over 200 additional resources and thousands of positive reviews are already available on the Civitai platform. Developers also announced two new versions — Z-Image-Base for refinement and Z-Image-Edit for image editing based on textual instructions, which will be available soon.
Z-Image-Turbo is already available for download and use , and many AI enthusiasts consider it the best solution for generating realistic images on regular home computers.
