Introducing GLM 5.3 on Amazon Bedrock

GLM 5.3 from Z.ai (Zhipu AI) is now available on Amazon Bedrock, the company announced. The model, published on Hugging Face Hub, is a 753B-parameter mixture-of-experts model optimized for coding and long-horizon agentic tasks. Access on Bedrock is available to eligible enterprise customers.

On Amazon Bedrock, GLM 5.3 is offered through fully managed APIs with cross-Region inference, prompt caching, and service tiers. It can be invoked through the OpenAI-compatible Responses and Chat Completions APIs, or the Amazon Bedrock Invoke and Converse APIs. Cross-Region inference is available through the US profile (us.zai.glm-5.3) and the Global profile (global.zai.glm-5.3). Service tiers include Flex for less-time-sensitive workloads, Priority for latency-critical requests at a higher price, and Standard.

Prompt caching is supported implicitly by default, with explicit cache controls recommended on the Responses and Chat Completions APIs. Each explicit cache breakpoint must contain at least 1,024 tokens to be eligible.

Z.ai claims competitive performance on coding benchmarks including DeepSWE, Terminal Bench 3.0, and FrontierSWE, and reports a 50% improvement over GLM 5.2 on its own internal coding benchmark. Direct comparisons to GLM 5 were not reported, because the magnitude of improvements led to updating the benchmark tests themselves since the GLM 5.1 announcement.

Z.ai also reported notable cyber security capabilities, measuring a leading score of 84.5 on the CyberGym benchmark at release. AWS described the model as a fit for defensive security workflows.

The post walks through invoking the model with the OpenAI Python SDK and the aws-bedrock-token-generator library, and demonstrates an authorized security test using Strix, an open-source AI penetration testing agent whose documentation currently uses GLM 5.3 as its default model. The walkthrough targets OWASP Juice Shop running locally. AWS noted that only applications one owns or has explicit written permission to test should be tested, and that unauthorized security testing is illegal in most jurisdictions and violates the AWS Acceptable Use Policy. AWS also noted that LiteLLM does not yet resolve bedrock/global.zai.glm-5.3, requiring the Converse API route and inference profile ARN to be specified explicitly.