Origin
Jie Tang founded Zhipu AI in Beijing in 2019 as a spin-out from Tsinghua University, according to Wikipedia. Its flagship product is the GLM family of language models. The name stands for General Language Model, and the training method behind it was introduced in a research paper in May 2022. Zhipu renamed itself Z.ai internationally in July 2025 and listed on the Hong Kong Stock Exchange on 8 January 2026.
How it got here
In April 2025 Zhipu disclosed that it was preparing to go public, and released new models under the MIT licence that month. GLM-4.5 and GLM-4.5-Air came in July 2025. The GLM-4.5 card says the company open-sourced the base models, the hybrid reasoning models and FP8 versions of those reasoning models. GLM-4.5V, a vision-language model, arrived in August 2025.
GLM-4.6V and GLM-4.7 came in December 2025, and a smaller GLM-4.7-Flash model also appears on the Hugging Face page. GLM-5 was released in February 2026. Wikipedia says GLM-5.1 went first to subscription users in late March 2026 and was released as open source on 7 April 2026. GLM-5.2 followed on 16 June 2026 with a one-million-token context window. GLM-5.3 arrived in August 2026, using the same 743 billion parameter base model as GLM-5.2 with improved post-training.
Licensing pattern
Z.ai has used the MIT licence for its open models since 2025. Wikipedia says its models have been released under MIT since July 2025, and the GLM-4.5 card says they can be used commercially and for secondary development. The model cards for GLM-5 and GLM-5.3-Flash also list MIT.
GLM-5.3 does not. Its licence is the custom GLM-5.3 License, which starts with MIT’s permissions: you may use, copy, modify, distribute, sublicense and sell the software, including the weights and code. Then it adds a condition for large model-hosting businesses. If you or your affiliates run a “Model as a Service” business and your combined revenue exceeds 10 billion US dollars over any 12 consecutive months, you must pass a security review by Z.AI before using the software or derivative works commercially.
The licence defines “Model as a Service” as giving a third party access to model inference or fine-tuning, for example through an API, in a way that gives that party meaningful control over inputs, parameters or training data. End-user products with model features built in are excluded.
That leaves GLM-5.3 and GLM-5.3-Flash, from the same lab, on different licences. Check the card for the specific model.