Z.ai (formerly Zhipu AI, a Chinese AI company) operates the GLM family of large language models. The current flagship GLM-5.2 is a Mixture-of-Experts model with approximately 744 billion total and around 40 billion active parameters, released as Open Source weights under MIT license. It offers a context window of up to 1 million Token and is specialized in agentic, long-term software engineering and coding tasks. Access is available via the Z.ai API, the GLM Coding Plan, or self-hosted through Hugging Face.

Preuves (2)