Chinese AI Models Near Parity with Leading US Counterparts

Chinese firm Z.ai, formerly known as Zhipu AI, is advancing artificial intelligence models that approach the performance levels of leading American counterparts, even surpassing models from Anthropic and OpenAI in certain respects.

The company concentrates on developing the open GLM series of models, which stand out for their ability to be deployed within the infrastructure of organizations possessing the necessary computing resources, enabling independent use without reliance on external services. This approach facilitates integration processes but also introduces security challenges.

In contrast, OpenAI’s models are accessible only through an application programming interface (API), imposing various restrictions, particularly for users in Russia.

The latest model in the GLM series, GLM-5.2, is close in capability to Claude Opus 4.8 and GPT-5.5. Despite a slight lag in some tasks compared to the newest OpenAI and Anthropic models, GLM-5.2 remains open and available via multiple service providers, including those operating in the Russian market.

Chinese models demonstrate strong results in certain areas; GLM-5.2 can outperform Western competitors in specific scenarios. However, Valentin Malikh, head of the Large Language Model Development Center at MWS AI, noted that benchmark tests only capture limited aspects of model performance and do not provide a comprehensive assessment of quality.

Malikh stated, “Based on practical experience, both personally and from colleagues’ trials, the Chinese model still trails slightly behind the advanced models from OpenAI and Anthropic. Whether the next generation of GLM will lead the field remains uncertain, but the pace of development in this sector makes this scenario possible.”

Challenges in Matching AI Market Leaders

Malikh added that catching up with AI market leaders remains difficult financially and regulatorily, given their vast capital and accumulated investments. He suggested that shifting the competitive balance would require a fundamental technological breakthrough capable of reshaping the market.

Dmitry Borozdin, co-founder and CEO of RetailCRM, commented that surpassing leading companies might necessitate adopting entirely new technological architectures, such as energy-efficient biological materials for neural connections or quantum processors. He emphasized that large investments alone are insufficient to overtake current market leaders.

Economic Efficiency as a Competitive Edge

Andrey Sharipov, product manager at Fedot.ai, part of the digital logistics operator versta.io, remarked that Z.ai’s surpassing of American models will remain challenging in terms of capabilities. However, he viewed competition as potentially decided on value for money, explaining that most users would be unwilling to pay significantly more for marginal performance improvements, which gives Chinese models a competitive advantage in economic efficiency.