Last updated:
GLM
Z.ai's open-weight GLM family — general-purpose reasoning and coding models from China.
Model Facts
Latest model: GLM-5.3
- Provider
- Z.ai (Zhipu AI)
- Released
- Aug 29, 2026
- Context window
- 200K tokens
- Weights
- Open
Full Model List
| Model | Released | Status | Context | Note |
|---|---|---|---|---|
| GLM-5.3 | Aug 29, 2026 | current | – | open release that outperforms some frontier models in vendor tests |
| GLM-5 | Feb 11, 2026 | legacy | 200K tokens | 744B-parameter MoE (40B active), 200K context, open weights |
- Official model page:
- z.ai
- Official documentation:
- docs.z.ai
- Other sources:
- GLM-5 (Hugging Face blog)
- GLM-5.3 (Gigazine)
What it is
GLM is the model family from Z.ai (Zhipu AI). GLM-5 (February 11, 2026) is a 744B-parameter mixture-of-experts model with 40B active parameters and a 200K-token context window; at launch it ranked as the top open-weight model on Artificial Analysis. GLM-5.3 followed on August 29, 2026 as an open model that Z.ai says outperforms some frontier models on selected tests.
Where this information comes from
Benchmarks shown are for GLM-5 from its launch coverage; no independent figures for GLM-5.3 are shown yet. Release dates, context windows and benchmark scores are vendor-reported unless stated, and harnesses differ between labs, so treat the numbers as indicative.
Notable traits
- Open weights on Hugging Face.
- Uses DeepSeek Sparse Attention for efficient long context.
- Strong coding and agent benchmark results for an open model.