Z.ai Releases 355B Parameter GLM-4.5 Under MIT License

Z.ai has introduced GLM-4.5, a powerful new open-source model designed for complex reasoning and agentic tasks. The model utilizes a Mixture-of-Experts (MoE) architecture with a total of 355 billion parameters, placing it firmly in the top tier of openly available large language models.

The most significant aspect of this release may be its license. GLM-4.5 is available under the permissive MIT license, a notable decision that removes the commercial use restrictions often seen with other flagship-class model releases. This choice allows developers and businesses to freely build and deploy applications on a state-of-the-art foundation.

Beyond Text Generation

Beyond its scale, GLM-4.5 is explicitly engineered for more than simple text completion. The model's architecture is optimized for hybrid reasoning, enabling it to function as the core of an autonomous agent that can understand user intent and execute multi-step tasks. Key capabilities include:

• Interacting with external tools and APIs.
• Planning and executing complex workflows.
• Performing sophisticated data analysis.

With a context length of 131,072 tokens, the model can process and reason over extensive documents. Developers can access the model weights and documentation on the official Hugging Face repository.