Z.ai GLM 5.3 Launches on Amazon Bedrock Coding and agentic workloads are asking more of AI models than ever. Refactor a repository spanning hundreds of files without losing context. Sustain a multi-hour workflow while reasoning through complex systems problems. Meeting these demands with open-weight models has historically meant provisioning your own inference infrastructure. GLM 5.3 from Z.ai is now available on Amazon Bedrock. You can use it through fully managed APIs without managing any infrastructure. Access to this model is available to eligible enterprise customers.
Model Specifications and Parameter Count GLM 5.3 is a mixture-of-experts model optimized for coding and long-horizon agentic tasks. As published on Hugging Face Hub, the model contains 753 billion parameters. This massive size supports complex reasoning and sustained context across large code bases. Z.ai has reported that the model shows notable cyber security capabilities. It builds on the same lineage as GLM 5, which arrived on Amazon Bedrock earlier this year. GLM 5.3 brings a range of important gains over its predecessors.
Coding Performance and Benchmark Gains Z.ai claims competitive performance on a range of coding benchmarks including DeepSWE, Terminal Bench 3.0, and FrontierSWE. They report a 50% improvement over GLM 5.2 on their own internal coding benchmark. Direct comparisons to GLM 5 were not reported because the magnitude of improvements led to updating the benchmark tests themselves. The updated tests reflect the significant jump in capability since the GLM 5.1 announcement. Stronger coding performance makes the model a natural fit for developer tools.
Emergent Cyber Security Capabilities Reported benchmark performance on security tasks stands out significantly for this new release. Z.ai measured a leading score of 84.5 on the CyberGym benchmark at release. This high score positions the model as a strong candidate for defensive security workflows. It can assist in automated security testing and vulnerability identification. The capabilities emerge naturally from the model's training and architecture design.
API Features and Cross-Region Inference GLM 5.3 supports cross-Region inference profiles for US and Global processing. You send requests to the source AWS Region of your choice, and Amazon Bedrock securely routes each request for processing. Implicit and explicit prompt caching are available features on the Responses and Chat Completions APIs. These APIs support both OpenAI-compatible interfaces and native Amazon Bedrock Invoke and Converse methods. Service tiers include Flex, Priority, and Standard options for different workload needs.