GLM 5.3

About this model

ZAI GLM 5.3 is a 744B parameter Mixture-of-Experts (MoE) language model built on the GLM-MoE DSA architecture, using the same base model as GLM-5.2 with significant post-training improvements focused on complex coding and long-horizon agentic tasks. It features a 1-million token context window and configurable reasoning effort levels (low, high, max), with thinking enabled by default. The model achieves strong benchmark results including 88.2 on Terminal Bench 2.1, 66.9 on DeepSWE (v1.1), 62.5 on HLE with Tools, and 73.0 on Toolathlon Verified. It is released under a custom license.

Technical specifications

Capabilities

Input modalities

Output modalities

Reasoning

Hybrid Default on

Knowledge horizon