Z-Image
Open-source 6-billion-parameter image generation model family from Tongyi that supports text-to-image, image editing, and fast sub-second inference.
About
Z-Image is a family of open-source image generation models developed by Tongyi-MAI. The foundation model focuses on high-quality, aesthetically rich images across diverse artistic styles and supports fine-tuning and negative prompting. Z-Image-Turbo, the distilled variant, achieves sub-second inference on enterprise GPUs using only 8 function evaluations and fits within 16 GB of VRAM on consumer hardware.
A third member, Z-Image-Edit, handles instruction-based image-to-image transformations for creative editing tasks. The models use a scalable single-stream diffusion transformer architecture and excel at photorealistic output with bilingual text rendering in English and Chinese. All models are released openly via Hugging Face and ModelScope at no cost.
A third member, Z-Image-Edit, handles instruction-based image-to-image transformations for creative editing tasks. The models use a scalable single-stream diffusion transformer architecture and excel at photorealistic output with bilingual text rendering in English and Chinese. All models are released openly via Hugging Face and ModelScope at no cost.