Zhipu AI releases GLM-5.3 with enhanced cybersecurity and coding capabilities

Zhipu AI (Z.ai) has released GLM-5.3, a new flagship large language model designed to enhance capabilities in coding, long-range agents, and cybersecurity. The model utilizes a Mixture-of-Experts architecture with approximately 744 billion parameters, activating about 40 billion per token.

Significant improvements have been noted in technical tasks. In cybersecurity evaluations, GLM-5.3 demonstrated an ability to develop multi-step reasoning chains to identify and plan exploit sequences, scoring 54.4% on ExploitBench, a substantial increase from its predecessor. In software engineering benchmarks like DeepSWE v1.1, the model's score rose from 46.2 to 66.9.

ZenMux has announced that GLM-5.3 is now available through its enterprise LLM platform, offering a free trial to allow developers to evaluate the model's API compatibility and operational fit within their existing workflows without needing to rebuild integrations.