GLM 5.3 Model Now Available on Amazon Bedrock

Z.ai's GLM 5.3, a 753 billion‑parameter mixture‑of‑experts model designed for coding and long‑horizon agentic tasks, is now accessible to eligible enterprise customers through Amazon Bedrock. The model is hosted on the Hugging Face Hub and can be invoked via fully managed APIs without any infrastructure to operate.

Compared with the earlier GLM 5 release, GLM 5.3 adds stronger coding performance—Z.ai reports competitive results on several coding benchmarks and a 50 % improvement on its internal benchmark—and emergent cyber‑security capabilities, including a leading score of 84.5 on the CyberGym benchmark. Bedrock integration provides OpenAI‑compatible Responses and Chat Completions APIs as well as Invoke and Converse APIs, implicit and explicit prompt caching, US and global cross‑region inference profiles, and three service tiers (Flex, Standard, Priority).

The blog demonstrates an agentic workflow using the open‑source Strix penetration‑testing agent, which can run security tests against a local application while leveraging GLM 5.3 on Bedrock. This example highlights how developers can experiment with managed inference, caching, and multi‑step tool‑augmented workflows directly from the Bedrock console or code.