Z.AI's GLM-5.3, a 744B-parameter MoE with 1M context, lands on Amazon Bedrock

Zai_org · x · 2026-10-06

Z.AI's flagship open-weight model GLM-5.3 is now available on Amazon Bedrock for eligible enterprise customers.

Key details:

Architecture: Mixture-of-Experts with 744B total parameters and 40B active per token, built on the same base as GLM 5.2 with improvements from scaled post-training.
Context: 1M-token context window with up to 128K output tokens, plus selectable reasoning effort levels to trade latency and token usage against performance.
Use cases: agentic coding, terminal/CLI task execution, repo-scale code generation and refactoring, long-horizon multi-step engineering, infrastructure diagnosis, and long-context codebase analysis.
Features: function calling, structured JSON output, and streaming responses including streamed reasoning content.
Launch date listed as October 5, 2026, with at least 45 days EOL notice.

More from Models