Chinese AI models are attracting more U.S. companies as businesses search for lower-cost artificial intelligence solutions. Rising expenses have encouraged developers and firms to compare alternatives while maintaining performance standards.
Several Chinese developers, including DeepSeek and Z.ai, now compete with advanced systems from OpenAI and Anthropic. Industry analysts say recent improvements have narrowed the performance gap between Chinese and American AI technologies. Businesses now evaluate models based on price, efficiency, and operational requirements instead of relying on one provider.
New data from OpenRouter highlights the growing adoption of Chinese AI models. The platform provides developers with access to multiple artificial intelligence systems through one service. Since February, more than 30% of weekly token usage by U.S. companies has gone to Chinese models. The figure reached (46%) during its highest recorded week.
The latest trend marks a major change from previous usage patterns. During the previous year, Chinese models averaged about 11% of usage. That share declined to 4.5% during the first half of 2025 before increasing sharply in recent months.
Experts link the growth mainly to lower operating costs. Kyle Chan of the Brookings Institution’s John L. Thornton China Center said companies have become more careful with AI spending as expenses continue increasing. Many businesses now seek better value from every AI deployment.
Open source and open weight systems have also supported the expansion of Chinese AI models. These models allow developers to inspect, modify, and deploy technology for different business needs. The added flexibility appeals to companies building customized AI applications.
AI startup Lindy shifted all of its traffic from Anthropic's Claude models to DeepSeek during June. The company estimated the move could save millions of dollars within months. Meanwhile, Z.ai's GLM 5.2 model also recorded rapid developer adoption after its launch.
Industry experts estimate some Chinese open source models cost 60% to 90% less than leading systems from OpenAI and Anthropic. Analysts also believe leading Chinese models remain only several months behind top American competitors. According to OpenRouter's Justin Summerville, newer open source systems can complete most AI tasks except the most complex workloads. For Pakistani businesses, these developments offer promising and important insights into future AI adoption strategies.
