AWS announced the release of GLM 5.3 on Amazon Bedrock, saying it is optimized for coding and long-horizon agentic tasks. It is Z.ai's first major update to its model series since the release of GLM 5 earlier this year.
Z.ai reported a leading score of 84.5 on the CyberGym benchmark, measured on security tasks. That compares with no prior benchmark scores provided by the company for earlier versions.
GLM 5.3 is built on the Hugging Face Hub and targets complex systems engineering and long-horizon agentic workflows. Availability begins for eligible enterprise customers, initially for those with access to Amazon Bedrock.
"GLM 5.3 is designed for complex systems engineering and long-horizon agentic tasks," said Z.ai. "These include multi-step reasoning, tool-augmented workflows, and sustained context across large code bases."
The announcement follows the release of GLM 5 on Amazon Bedrock, which was framed as a foundational step in expanding AI capabilities for enterprise use. AWS did not say how GLM 5.3 compares to GLM 5 directly, and noted the model's potential for cybersecurity applications.
Source: awsml
