#AI Innovation #Z.ai #GLM-4.5

Z.ai Unveils Next-Generation AI Model: GLM-4.5

In a groundbreaking announcement dated July 28, 2025, Z.ai, a leading Chinese AI enterprise formerly recognized as Zhipu, introduced its latest model, GLM-4.5. This innovative open-source model is designed to redefine the standards of performance and accessibility in the field of artificial intelligence. Built on a fully proprietary architecture, GLM-4.5 is recognized as the most cutting-edge open-source Mixture of Experts (MoE) model in China, and it is poised to offer developers an affordable AI solution without compromising on quality.

A Technological Milestone

The GLM-4.5 model series includes two distinct versions: the robust flagship GLM-4.5, featuring an impressive 355 billion total parameters, and the more streamlined GLM-4.5-Air, with a conservative 106 billion parameters. This model series illustrates Z.ai's commitment to achieving SOTA (state-of-the-art) performance while maintaining parameter efficiency. In test benchmarks assessing its capabilities in reasoning, coding, and other agentic functions, GLM-4.5 has earned a noteworthy global ranking, being second overall and first among both domestic and open-source models.

Notably, GLM-4.5 has demonstrated exceptional generation speed, attributed to its ability to generate high-quality text at a rate exceeding 100 tokens per second. Furthermore, it provides an economical pricing structure for API usage, costing $0.11 per million input tokens and $0.28 per million output tokens. These features fundamentally enhance both developer experience and user accessibility in utilizing AI technologies.

Native Agent Capabilities

One of the most striking aspects of GLM-4.5 is its