Zhipu AI Unveils Open-Source GLM-4.5 Language Model

Zhipu AI, a Chinese artificial intelligence company previously recognized by OpenAI as a significant global competitor, has introduced its latest language model, GLM-4.5, which is now open-source. Released on Monday evening, this model serves as a foundational framework for various agent applications and has showcased advancements in technology that allow for sophisticated reasoning, code generation, and interactive capabilities.

According to the company, GLM-4.5 stands out as the first native agent model of its kind, achieving the top rank in domestic evaluations, while displaying performance metrics comparable to leading global models. The modeled efficiency in terms of parameters has seen a twofold increase, and the API pricing is remarkably lower—only one-tenth the cost of the Claude model, with the capacity to process over 100 tokens per second.

The GLM-4.5 model uniquely amalgamates several critical functions, including reasoning, coding, and agent abilities into a single framework for the first time. This integration renders it adept for a variety of applications, from generating dialogue to performing complex agent tasks, thereby broadening its usability in the AI landscape.