Amazon Bedrock adds Z.ai's GLM 5.3 for enterprise coding workloads

Amazon Web Services made Z.ai's GLM 5.3 available through Bedrock on October 5. Access is limited to qualifying enterprise customers, with the model aimed at programming and extended software engineering tasks.

GLM 5.3 contains 753 billion parameters and activates approximately 40 billion for each token it processes. It supports a context capacity of one million tokens and responses of up to 128,000 tokens. Reasoning remains enabled, with adjustable effort settings.

The Bedrock integration lets businesses use the open-weight model without operating its underlying computing infrastructure. It also supports caching repeated prompt content, which AWS says reduces processing delays and input costs. That feature applies to workflows that repeatedly submit the same instructions or code.

Eligible customers can access GLM 5.3 through Bedrock's console or supported APIs. Requests can use either US or global inference profiles, which route processing across AWS regions.