GLM on Berges AI
The open-weights model family from Z.ai, a large reasoning model for the hardest questions. Pick it from the sidebar and it answers every turn.
GLM is a family of open-weights large language models from Z.ai, the Beijing lab formerly known as Zhipu AI. The line grew out of Tsinghua University research and has become one of the leading open-weights families, with a focus on reasoning, coding, and tool use.
Berges AI serves the current GLM model in its deep-thinking layer. It is one of the deeper options, so it goes further on hard questions than a quick model and costs more units per reply. Pick GLM from the sidebar and it answers every turn. Your choice is remembered.
The weights are published openly, so anyone can audit them, fine-tune them, or host them. The license is MIT-style for almost everyone; only the very largest cloud providers must pass a review by Z.ai before offering it commercially.
GLM specs
Maker: Z.ai, formerly Zhipu AI
License: GLM license, MIT-style (open weights)
Architecture: Mixture-of-Experts
Current version: GLM 5.3
Best known for: Reasoning, coding, and tool use
What GLM is good at
Depth on hard problems
GLM puts more of its parameters to work on every word than a quick model does. On questions with several dependent steps, it holds the thread and checks its own work more reliably.
Code
Z.ai has trained the GLM line heavily on programming and agent work. Reading unfamiliar code, planning a change, and debugging are where it is strongest.
Bilingual roots
GLM was built for Chinese and English from the start, and it holds up well across other major languages.
When GLM is the right pick
- Hard reasoning where a quick answer feels shallow
- Reading, planning, and debugging code
- Analysis with several dependent steps
Things people ask about GLM
Can I use GLM without Berges AI?
Yes. Z.ai publishes the weights openly. You can run them yourself or use any open-weights host. Berges AI is one place to chat with GLM, with privacy as a default and encryption at rest.
When should I pick GLM?
When the question is hard: code, multi-step analysis, or a problem where a quick answer feels shallow. GLM is one of the deeper models, so each reply costs more of your plan's units than the quick one.
Is GLM really open source?
Open weights, under Z.ai's own license. It lets you use, modify, and sell products built on the model, like MIT. The one extra condition applies only to cloud providers with more than $10 billion in yearly revenue, who need a review by Z.ai before hosting it commercially.