Z-image AI

Z-Image AI is a next - generation efficient image generation foundation model launched by Alibaba's Tongyi team in November 2025. It relies on the innovative S³-DiT (Scalable Single-Stream Diffusion Transformer) architecture and stands out in the field of image generation with the advantages of small parameters, high performance and low power consumption. Here is a detailed introduction to it:
Core Technical Advantages
Innovative single - stream architecture: It breaks the information barrier of the traditional dual - stream method.

It connects text semantics, visual semantics and image tokens into a single data stream for processing. With only 6B parameters, it achieves performance comparable to or even better than that of top - tier models. Meanwhile, it cuts the parameter quantity to one - third of that of competitors and significantly improves inference efficiency. Fast inference and hardware friendliness: