GLM-5.2 was released as an open-weight model by the PRC-based company Z.ai (formerly known as Zhipu AI) on June 16, 2026, and CAISI completed its assessment on July 8.

In CAISI's full assessment of GLM-5.2

, the key findings were that:

 

  • GLM-5.2 was probably the most capable open-weight AI model when it was released (Figure 1). According to CAISI’s set of evaluations:

    • GLM-5.2’s overall capabilities are similar to that of GPT-5.2, released in December 2025.

    • GLM-5.2’s cyber capabilities are similar to that of Opus 4.6, released in February 2026.

       

  • GLM-5.2’s performance on safeguards and security is mixed. According to CAISI’s set of evaluations:

    • GLM-5.2’s safeguards allow assistance with agentic cyber exploit development.
    • GLM-5.2’s safeguards block fewer sensitive biological questions than reference U.S. models.
    • However, GLM-5.2 appears potentially more robust against agent hijacking and jailbreaking attacks than other evaluated PRC open-weight models.
    • These evaluations measure robustness against prompt-based jailbreaks. Regardless of their robustness, safeguards for open-weight models can be circumvented when self-hosted.
Figure 1: Comparison of aggregate capabilities over time of the most capable U.S. and PRC models as of GLM-5.2’s release. A 400-point increase on the y-axis equates to a 10x increase in the odds of solving tasks. Open circles denote non-frontier models, and error bars and shaded regions denote 95% CIs. For details about the evaluations and methodology used to generate this chart, see Appendix A1 and Appendix A4 of the full assessment.

Credit: NIST