GLM-Image

GLM-Image is a state-of-the-art open-source AI image generation model developed by Z.AI (Zhipu AI), giving creators, designers, and developers a powerful tool they can freely build on under the permissive Apache-2.0 license.

What sets GLM-Image apart is its groundbreaking hybrid architecture: a 9B-parameter autoregressive transformer works in tandem with a 7B-parameter diffusion decoder. This design delivers industry-leading performance where most image models struggle — precise text rendering inside images, scoring 0.9116 accuracy on the CVTG-2K benchmark — as well as knowledge-intensive content generation, so posters, diagrams, and infographics come out with legible, correctly spelled text and accurate detail.

GLM-Image supports high resolutions up to 2048×2048, making it suitable for print-quality assets and detailed compositions. The glm-image.top site operates as an independent service that deploys and optimizes these open-source models, offering a convenient way to try the technology without setting anything up yourself.