Amazon Bedrock now offers GLM 5.3 from Z.ai. It is a 753B-parameter mixture-of-experts model. The model targets coding and long-horizon agentic tasks. Z.ai reports notable cyber security capabilities. It is accessible via fully managed APIs with cross-Region inference. Prompt caching is supported to reduce latency and cost. Service tiers let users balance cost and speed. The model can be invoked through OpenAI-compatible APIs. It is also available via Bedrock’s Invoke and Converse APIs. GLM 5.3 shows a 50% improvement on Z.ai’s internal coding benchmark. It scored 84.5 on the CyberGym benchmark. Cross-Region inference profiles are available for US and global routing. Explicit prompt caching can be enabled for better cache hit rates.
What changed
GLM 5.2 was the latest version available on Amazon Bedrock.
What it unlocks
Running long-horizon agentic workflows with sustained context and reduced latency via prompt caching.
- 753B parameters
- 84.5 on CyberGym benchmark
- 50% improvement on internal coding benchmark
What you need to act on it
- access to Amazon Bedrock
- eligible enterprise customer status
Sources