GLM 5.3 Arrives on Amazon Bedrock: 1M-Token Context for Long-Horizon Coding
Hi, it's Shii-chan! A new open-weight model just joined Bedrock.
AWS What's New
aws.amazon.com
What was announced?
In AWS What's New, GLM 5.3 from Z.ai was announced as generally available on Amazon Bedrock. It is presented as an option for agentic coding and long-horizon software engineering work.
The highlight is that you get the latest open-weight model within the AWS security and compliance posture.
Why it matters
GLM 5.3 is Z.ai's flagship model, built on the same base model as GLM 5.2, with improvements driven by scaled post-training. You can now pick a long-context coding model right inside Bedrock.
What changes
- You can call GLM 5.3 from the Bedrock console or supported APIs
- You can place cache points on system prompts and messages to cut latency and input costs when reusing context
- You can choose a reasoning effort level to trade latency and token consumption against task performance
Dive Deep
Here are the specs from the original post.
- Architecture: Mixture-of-Experts, 753B total parameters, roughly 40B active per token
- Context window: 1 million tokens
- Output: up to 128K tokens
- Reasoning: always enabled, with selectable effort levels
- Prompt caching: explicit caching with cache points on system prompts and messages
Keep the availability conditions in mind too.
- Available to eligible enterprise customers
- Accessed through the US and Global cross-Region inference profiles
For supported Regions, endpoints, features, and pricing, see the Amazon Bedrock documentation.
Wrap-up
- Z.ai's GLM 5.3 is GA on Amazon Bedrock
- 753B MoE (about 40B active), 1M-token context, up to 128K output
- Reasoning is always on with selectable effort, plus explicit prompt caching
- Eligible enterprise customers can use it via the US and Global inference profiles
This is for anyone building coding agents on Bedrock that take on long tasks!
