Z.ai's GLM-5.3 Now Accessible via API for Developers
Chinese startup z.ai has launched GLM-5.3, its latest open-source language model, on its API, enabling developers to integrate it into applications and agents. The model gained attention last week for reportedly uncovering a previously unknown vulnerability in Cursor, showcasing advanced cyber capabilities. Pricing is set at $1.4 per million input tokens and $4.4 per million output tokens, with open-weight release planned but not yet dated.
Z.ai has opened access to its GLM-5.3 language model through a developer API, allowing teams to embed the system into their own products and automated workflows. The model carries an open-source designation, which positions it as an alternative to proprietary offerings in the rapidly expanding AI landscape.
The model attracted significant industry attention last week after reportedly detecting a previously unknown security flaw in Cursor, a widely used coding assistant. Developers can now access the model at $1.4 per million input tokens and $4.4 per million output tokens. An open-weight release has been confirmed as planned, though no specific date has been provided.
GLM-5.3's demonstrated ability to uncover real-world software vulnerabilities could accelerate the use of AI in security auditing, potentially helping developers patch flaws faster. However, the same capability may also lower the barrier for malicious actors seeking to exploit weaknesses, raising questions about responsible deployment. The competitive pricing could pressure other model providers, possibly making advanced AI more affordable for smaller teams while intensifying scrutiny of open-weight releases.