Zhipu GLM-4.6V
Cherry Studio users can now use the built-in CherryIN free trial of the service Zhipu GLM-4.6V— a vision flagship model released by Z.ai (Zhipu AI) in December 2025, with a MoE architecture, 128K native multimodal context, and native multimodal tool calling, making it the top choice for image-text understanding and multimodal Agent scenarios.
🚀 What is GLM-4.6V?
GLM-4.6V is the latest-generation vision-language model in Z.ai's GLM-V series, natively supporting unified text + image modeling and further expanding context and tool-calling capabilities on the basis of GLM-4.5V.
Architecture: Mixture-of-Experts (MoE)
Total parameters: 106B
Active parameters: about 12B
Context length: 128K tokens
Open-source license: MIT
Release date: December 8–9, 2025
Visual encoder: supports multi-resolution images (up to 4K)
The series also includes GLM-4.6V-Flash (9B), designed for local and low-latency scenarios, free for commercial use.
📚 Continuing the multimodal training system of the GLM-V series
GLM-4.6V follows the technical approach of GLM-4.1V-Thinking / GLM-4.5V and further strengthens vision and Agent capabilities:
Native multimodal modeling: joint training of text and images, supporting mixed text-image input
Context expansion: training context extended to 128K tokens, able to process about 150 pages of dense documents, 200 pages of slides, or 1 hour of video in a single pass
Native multimodal tool calling: tools can directly receive and return images, processing multimodal artifacts via URL based on the extended MCP protocol
Reinforcement learning enhancement: continues the scalable RL process of the GLM-V series
⚙️ Native multimodal, designed for real-world scenarios
GLM-4.6V's multimodal capabilities cover everyday and professional scenarios:
✅ Rich text content understanding: long documents, multi-page text, and mixed text-image layouts
✅ Visual web search: online retrieval and understanding combined with visual input
✅ Frontend recreation: generate frontend code from design drafts or UI screenshots
✅ Long-context multimodal document analysis: whole PDF / slides / video-level input
✅ Chart and table parsing: structured information extraction
💡 Native multimodal tool calling and Agent capabilities
One of the core upgrades of GLM-4.6V is "visual perception → executable action" the closed loop: tool calling natively supports images as both input and output, enabling multimodal Agents to be deployed in real business scenarios.
Simple image-text Q&A
Direct conversation
"What’s in this picture?"
Moderately complex tasks
Enable tool calling
Retrieve data after reading charts
Complex multimodal Agent
Multiple tools + MCP
Screenshot → understanding → API call → report generation
🌟 Efficient MoE, open and available
⚡ MoE sparse activation: 106B total parameters, only about 12B activated
💰 Through CherryIN in Cherry Studiofree to use
🖥️ Weights, inference code, and MCP tools are open-sourced on GitHub and Hugging Face, under the MIT license
🧠 Focused on practical capabilities: multimodal assistant
GLM-4.6V is suitable for the following scenarios in actual use:
Document assistant: read and summarize full-length documents, scans, and slides
Data analysis: identify and interpret charts and dashboard screenshots
Frontend and design: generate or modify frontend code based on UI screenshots
Visual search: combine images for online retrieval and information integration
Multimodal Agent: complete complex tasks with tools such as browser, code execution, and retrieval
🧭 How to use it in Cherry Studio?
Open Cherry Studio and go to Settings → Model Services.
find CherryIN the service provider and enable it.
In the model list, select Zhipu GLM-4.6V.
Return to the chat screen and switch to GLM-4.6V, you can directly upload images in the conversation for image-text interaction.
💡 Tip: The free model quota provided by CherryIN is borne by the official Cherry Studio, suitable for daily experience and evaluation; for production environments, it is recommended to use the official Z.ai (Zhipu) API.
📘 Experience Zhipu GLM-4.6V now and unlock native multimodal and visual Agent capabilities!
💡 Get help and submit feedback
If you encounter any questions, bugs, or have suggestions for feature improvements during configuration or use, please refer to Feedback and Suggestions the official channels provided there.
Last updated
Was this helpful?