I have started paying more attention to companies outside the US, to prevent myself from getting the "frog in the well" syndrome. You spend ten years reading about OpenAI, Anthropic, and Google, and you begin to think that is the entire AI market. It is not.
I saw the news that Zhipu AI—also known as Z.ai—released GLM-5 on 11 February 2026. What struck me was not the model itself, but the business decision that followed: they raised prices immediately. The stock market rewarded them. This is what market validation looks like.
Who Is Zhipu AI
Zhipu was founded in 2019 by researchers from Tsinghua University. The mission was explicit: build frontier artificial intelligence using Chinese technology and Chinese infrastructure. This was not a startup chasing venture capital. This was a decision by Beijing to ensure China could develop AI without asking the US for permission.
For the past seven years, Zhipu has built the GLM series of models. They are not novel for technical reasons alone. They are novel because they exist: a world-class language model trained on Chinese chips, using Chinese frameworks, deployable entirely within Chinese infrastructure.
What GLM-5 Actually Is
GLM-5 is a 744-billion-parameter model trained entirely on Huawei Ascend chips. The technical report was submitted to arXiv on 17 February 2026. The model is open-source under MIT licence, available on Hugging Face and ModelScope, and free to test at chat.z.ai.
You can use it right now. You should also be careful when you do.
A Necessary Caution
There is active discussion in Western security circles about not downloading Chinese open-source language models unless you know what you are doing and can assure yourself that nothing is being transferred to the cloud. The risk is not paranoia. It is rational caution. If you test GLM-5 at chat.z.ai, do not use proprietary information. If you run the model locally, you control the data. If you do not know the difference between these two options, do not use the model yet.
Why Open Source Matters Here
Zhipu open-sourced GLM-5 under MIT licence. This is not charity. It is strategic alignment. When you release a frontier model under an open licence, you signal to governments and enterprises: you control this technology. You own the weights. You run it on your hardware. No vendor lock-in. No data flowing to Beijing. No dependence on American cloud infrastructure.
This is the entire point. Governments in Southeast Asia, the Middle East, Africa, and Latin America—regions where the US has sanctions or restrictions—can now deploy frontier AI without asking Washington for permission. They can audit the model. They can run it locally. They can know, with certainty, that their data stays in their infrastructure.
This is what "technology sovereignty" means in practice.
The Pattern
Zhipu is not alone. DeepSeek released open-source models. Other Chinese AI companies are moving in the same direction: build capability, open-source it, make it available globally, but with the understanding that this benefits the Chinese technology ecosystem first.
This is a quiet decoupling. Not hostile. Not closed. But deliberate: China is building AI infrastructure that does not depend on US semiconductors, US software frameworks, or US cloud platforms. And they are releasing it openly so other countries can do the same.
Why This Space Matters to Watch
If you only read about OpenAI, Anthropic, and Google, you miss half the story. The other half is being written in Beijing, Hangzhou, and Shanghai. It is written in open-source releases. It is written in stock prices rising 34 percent when a company raises prices. It is written in partnerships with ASEAN nations and Belt and Road countries that want AI infrastructure they control.
The US leads in proprietary frontier models. China is winning the infrastructure game. These are not the same thing. Understanding both is necessary to understand where AI is actually going.
Sources
- GLM-5 Team. "GLM-5: from Vibe Coding to Agentic Engineering." arXiv, 17 Feb. 2026, arxiv.org/abs/2602.15763.
- Z.ai. "GLM Coding Plan Price Adjustment." X/Twitter, 11 Feb. 2026.
- TrendForce. "Rising Costs and Demand Drive China's LLM Price Jump: Zhipu GLM-5 Hikes 30% in First 2026 Increase." TrendForce, 16 Feb. 2026.
- OpenRouter. "GLM-5 – AI Coding Model Performance & Pricing." OpenRouter, 2026, openrouter.ai/z-ai/glm-5.
- Hugging Face. "GLM-5 Model Weights." Hugging Face, 2026, huggingface.co/zai-org/GLM-5.
