China’s artificial intelligence startup Z.ai has unveiled its latest model, GLM-4.5, which undercuts its domestic rival DeepSeek in both model size and operational costs. With just eight Nvidia H20 chips required and agentic task-solving architecture, the model marks a pivotal moment in China’s effort to create efficient, cost-effective AI alternatives in the wake of U.S. chip restrictions.
China is accelerating its competitive edge in artificial intelligence with the launch of the new GLM-4.5 model by Z.ai. Previously known as Zhipu, the company has introduced an AI model that not only challenges local rival DeepSeek on functionality but also delivers a major cost advantage—underscoring China’s strategic pivot toward efficient, low-cost AI systems amid tightening global regulations.
The GLM-4.5 model is built on an “agentic” AI framework, a system designed to deconstruct complex tasks into smaller, manageable segments to improve accuracy and speed. In contrast to traditional transformer-based models, this design structure offers higher problem-solving efficacy and operational optimization.
Also Read: Baidu vs. DeepSeek: The Battle for China’s AI Crown
Notably, GLM-4.5 operates with just eight Nvidia H20 chips—customized GPUs approved under U.S. export compliance rules—representing a 50% reduction in model size compared to DeepSeek’s architecture. This design significantly reduces hardware dependency and supports the broader national goal of semiconductor self-reliance.
From a cost perspective, Z.ai has priced the model at $0.11 per million input tokens and $0.28 per million output tokens, substantially undercutting DeepSeek’s comparable rates. This pricing strategy is expected to attract startups, developers, and large-scale enterprises focused on affordability and scalability in AI deployment.
Z.ai has refrained from disclosing its model training expenditure but indicated that it has sufficient GPU resources to meet current demand. Industry watchers believe that this release is part of a broader push to assert China’s technological autonomy, especially as global scrutiny around AI capabilities and data governance tightens.
Founded in 2019, Z.ai is reportedly preparing for a public listing in Greater China and has raised over $1.5 billion from a robust portfolio of investors, including municipal innovation funds and sovereign venture capital. The company’s financing strength further validates its role as a critical player in China’s AI strategy.
Also Read: DeepSeek’s AI Breakthrough Signals China’s Chipmaking Ambition
Z.ai’s launch follows a growing trend in China toward open-source AI models. By enabling free developer access, Z.ai is betting on ecosystem growth and faster adoption rates. The strategic timing also coincides with a partial resumption of chip supplies to China, potentially easing future scale-up efforts.
With increasing geopolitical and regulatory uncertainty, Z.ai’s GLM-4.5 provides a compelling alternative for businesses looking to balance performance, compliance, and cost-effectiveness in their AI infrastructure.
As global competition heats up, China’s Z.ai is not just building models—it is redefining the economic equation of artificial intelligence.
READ MORE ON