US developers are adopting Z.ai’s low-cost GLM coding AI models

In short: More US developers are trying Z.ai’s GLM-4.7 and GLM-5.x AI models because they can help with coding tasks and cost much less than many US options.

What's going on

Z.ai, a Chinese AI company that was spun out of Tsinghua University, has been drawing attention in the US for its GLM family of AI models. These are large language models, which are systems that can read and write text and code, a bit like a very advanced autocomplete.

Since July 2025, Z.ai has released its GLM models under the MIT license. That is a common open-source license that lets people use and modify the software with few restrictions. In practice, this can mean a company can run the model on its own computers instead of paying for every use.

Developers say the models are strong at coding and “agentic” work, which means carrying out a task in multiple steps using tools (like an assistant that can plan, then execute, then check its own work). Benchmarks and reports cited in recent coverage say GLM-4.7 and GLM-5.x are competitive with top US models on coding and tool-use tests, and GLM-5.2 offers a very large context window, up to 1 million tokens (think of it as a very large working memory that can hold more of your code and notes at once).

Cost is a major reason for the interest. Reporting notes that some GLM options are free to use if you self-host, and paid plans and APIs can still be cheaper than leading US services.

What to watch

Watch whether more US startups and cloud providers offer GLM models as a standard option, and how companies balance lower cost with other concerns like security, safety rules, and support.

Source: NYTimes