ChatGLM3 is the third-generation open bilingual chat LLM series from Zhipu AI, headlined by the ChatGLM3-6B model. It was one of the most widely adopted open Chinese-English chat models of its generation, with stronger reasoning, cleaner formatting and native agent capabilities than its predecessors, and the repository has gathered about 13,600 stars on GitHub.
The series ships a base model and a chat model at the 6B scale, trained with a new prompt format and substantially more diverse data than ChatGLM2. Unlike earlier generations, ChatGLM3 was designed from the start to act as an agent: it natively supports function calling, a code interpreter and tool use inside multi-turn conversations, so the chat model can plan steps and invoke external tools rather than only producing text. A long-context variant extends the window to 128K tokens. The models run on consumer hardware, with quantized versions fitting on a single modest GPU, and the project provides deployment and fine-tuning code alongside the weights. The repository was last updated in 2025 and remains a reference point for open bilingual models.
ChatGLM3 is free and open source under the Apache-2.0 license. The weights can be downloaded and self-hosted at no cost; Zhipu separately sells API access to its newer hosted GLM models, but using this repository costs nothing beyond your own hardware.
ChatGLM3 is for developers and researchers building bilingual chatbots and agent prototypes, especially for Chinese-language applications, who want a proven open model they can fine-tune and run entirely on their own infrastructure.