Zhipu AI Releases GLM-4.7-Flash MoE Model

Chinese AI firm Zhipu AI has released GLM-4.7-Flash, a new language model designed for high-speed inference. It employs a Mixture-of-Experts (MoE) architecture, a technique that allows models to scale up their parameter counts while keeping computational costs manageable during inference.

The 'Flash' in its name signals the model's primary goal: performance. MoE models achieve this by selectively activating only a fraction of their total parameters—the 'experts'—to process any given input. This makes them significantly faster and more efficient for real-time applications compared to dense models of a similar size.

A Permissive License for Commercial Use

Perhaps most notably for developers and businesses, GLM-4.7-Flash is available under the MIT license. This is one of the most permissive open-source licenses, imposing very few restrictions on reuse and allowing for broad commercial applications. This combination of an efficient architecture and a business-friendly license makes the model an attractive option for integration into products and services.

The model is the latest addition to Zhipu AI's GLM-4 family of models. While specific details on its total parameter count and context length have not been disclosed, developers can access the model weights and resources on its official Hugging Face repository.