GLM Image is Z.AI's open-source AI image generation model featuring a hybrid autoregressive and diffusion architecture, designed for precise text rendering and knowledge-intensive content creation.
What is GLM Image?
GLM Image is a next-generation AI image generation model developed by Z.AI that combines a 9B autoregressive model with a 7B DiT diffusion decoder. It takes text prompts and optional parameters (resolution, aspect ratio) as input and outputs high-quality images with superior text rendering. The model is available via API for industrial-grade generation and also open-source for self-hosting.
Key Features
- Hybrid Architecture — Combines a 9B autoregressive model for semantic understanding with a 7B DiT diffusion decoder for detail portrayal, achieving balanced image quality.
- Precise Text Rendering — Achieves Word Accuracy of 0.9116 and NED of 0.9557, setting open-source SOTA for rendering text in images.
- Multiple Resolutions — Supports aspect ratios 1:1, 3:4, 4:3, and 16:9 with resolution range from 512px to 2048px.
- Knowledge-Intensive Scenarios — Excels at generating commercial posters, PPTs, and popular science illustrations that require complex layouts and detailed text.
- Fast Generation — Quick API calls suitable for industrial-scale image generation.
- Open Source — Model weights and code are publicly available, allowing customization and deployment.
- Freemium Access — Free tier available with usage limits; paid plans for higher throughput.
Who is it for?
- Graphic designers needing precise text embedding in images for posters, banners, and marketing materials.
- Content creators for social media and marketing who require fast, high-quality visuals with readable text.
- Educators creating educational materials such as infographics and science illustrations.
- Developers integrating AI image generation into applications via API or self-hosted deployment.
What can you do with GLM Image?
- Design commercial posters with accurately rendered product names and headlines, outputs in multiple supported resolutions.
- Generate social media visuals with custom text overlays, optimized for 1:1 or 4:3 aspect ratios.
- Create presentation slides (PPTs) with embedded charts and labels, leveraging knowledge-intensive generation.
- Produce educational infographics that require precise text and complex visual relationships.
How does GLM Image work?
The model operates in two stages: first, the autoregressive component generates a semantic layout and text content. Then, the diffusion decoder refines details and produces the final high-resolution image. The hybrid design allows end-to-end generation from a text prompt without manual layout steps.
Pricing
GLM Image follows a freemium model with a free tier offering limited API calls per month. Paid tiers provide higher usage limits and priority processing. The open-source version is free to use and modify.
FAQ
Is GLM Image free?
Yes, GLM Image is open source and also offers a free API tier for limited use. For high-volume production, paid plans are available.
What resolutions does GLM Image support?
It supports resolutions from 512px to 2048px with aspect ratios 1:1, 3:4, 4:3, and 16:9.
How accurate is the text rendering?
GLM Image achieves open-source SOTA with Word Accuracy of 0.9116 and NED of 0.9557, making it suitable for scenarios where text must be clear and readable.
Can I run GLM Image on my own hardware?
Yes, the model is open source and can be deployed locally or on your own infrastructure for full control and privacy.
What languages does GLM Image support for text in images?
It supports English and Chinese text rendering, with optimization for both languages.
