China’s Z.ai Unveils GLM-5.2, a 1M-Token Open Model

Z.ai released GLM-5.2 on June 13, 2026. The open-weight model targets long coding jobs and software engineering tasks. Z.ai lists the model at about 744 billion total parameters and uses a mixture-of-experts architecture with roughly 40 billion parameters active per token. The model supports a 1 million-token context window and is released under an MIT license with no regional limits.

Z.ai founder Jie Tang framed the release in a launch statement, writing, “Science should be global. The path to AGI must never be enclosed by high walls.” The announcement arrived the same week U.S. authorities ordered restrictions on foreign access to certain advanced models.

In a mixture-of-experts design, only a subset of the model’s parameters are used for each token. Z.ai’s configuration aims to provide high overall capacity while keeping compute per query lower than a fully active model. The company reports about 40 billion active parameters per token.

The 1 million-token window is about five times the limit of GLM-5.1. That larger window lets the model retain more of a codebase, documentation set or project history within a single session.

The MIT license permits downloading, running and modifying the model without regional restrictions. Z.ai positions GLM-5.2 for long-horizon coding tasks, complex engineering work and AI agents that must handle large information sets.

On Z.ai’s benchmark table, GLM-5.2 trails one competitor by less than one percentage point on the FrontierSWE coding test while outperforming another competitor on the same long-horizon evaluation. Vercel CEO Guillermo Rauch wrote on X that he was “genuinely impressed, almost shocked” by the model’s coding ability. Developers interested in open models responded quickly to the release.

Analysts are watching adoption costs and deployment issues. Lian Jye Su, chief analyst at Omdia, noted that enterprise buyers judge models on performance versus competitors and cost of adoption. Tulika Sheel, senior vice-president at Kadence International, warned that governance and operational stability will matter in real deployments.

Companies considering GLM-5.2 will evaluate how much infrastructure is required to host a mixture-of-experts model, whether the model maintains consistency and safety in production, and how to meet compliance requirements in regulated industries. The MIT license simplifies legal reuse, but firms must still validate reliability, content safety and alignment with internal policies before broad deployment.

For researchers and engineering teams, the open-weight release provides a base for fine-tuning, custom adaptations and independent audits. GLM-5.2 adds to the number of open models available for download and self-hosting and is available for companies and developers to test and deploy under the MIT terms.