GLM-5.2

A Baidu Cloud technical analysis describes GLM-5.2 as a next-generation open-source coding model built on the GLM-5 series architecture, with systematic optimizations for complex engineering scenarios, long-context agent tasks, and computational efficiency. The article reports 744 billion total parameters with approxim DSA uses an importance evaluation module scoring tokens by syntax role, contextual relevance, and task correlation, combined with dynamic attention allocation and gradient propagation through differentiable sparse gating. These design choices target the n² attention bottleneck for long-context coding and agent workload

The NIST Center for AI Standards and Innovation (CAISI) published an official assessment of Z.ai's GLM-5.2 on July 8, 2026, evaluating the open-weight model released on June 16, 2026. CAISI concluded that GLM-5.2 was probably the most capable open-weight AI model at release, with overall capabilities similar to GPT-5.2 On safeguards and security, CAISI found mixed results: GLM-5.2's safeguards allow assistance with agentic cyber exploit development and block fewer sensitive biological questions than reference U.S. models, though it appears potentially more robust against agent hijacking and jailbreaking than other evaluated PRC open-

Semgrep reports that Z.ai's open-weight GLM-5.2 outperformed Anthropic's Claude Opus 4.8 on their IDOR benchmark, a dataset and prompt Semgrep uses to evaluate frontier coding agents. Writing in the "We have Mythos at Home" post, authors Katie Paxton-Fear, Seth Jaksik, Brenden Noblitt, and Erik Buchanan noted surprise The benchmark results underscore GLM-5.2's practical impact for security engineering teams evaluating open-weight alternatives to proprietary coding agents. By surpassing Claude Opus 4.8 on a well-known cyber benchmark, GLM-5.2 demonstrates that self-hostable models can compete with top-tier closed systems on real vuln