Chinese AI models are rapidly gaining traction among American companies, driven by a combination of high performance and significantly lower costs. As the price of using advanced models from OpenAI and Anthropic continues to climb, businesses are increasingly turning to Chinese alternatives — DeepSeek, Z.ai (GLM 5.2), and Alibaba Qwen. This trend is reported by U.S. news outlet CNBC.
According to data from OpenRouter, the share of tokens attributed to Chinese models among American companies has consistently exceeded 30% since February, occasionally reaching 46% in certain weeks. For comparison, this average stood at just 11% over the previous 12 months.
Experts point to cost as the primary driver of this growth. Chinese models can be 60–90% cheaper than their leading American counterparts, while trailing the most advanced systems from OpenAI and Anthropic by roughly 6–9 months in terms of quality. For most corporate tasks, however, this level of performance is already sufficient.
Some companies have fully transitioned to Chinese models. For instance, the startup Lindy has shifted all of its traffic from Anthropic’s Claude to DeepSeek. The company estimates this move will save millions of dollars over the course of several months, while also improving results across key use cases.
New models are also gaining momentum quickly. Z.ai’s GLM 5.2 became the fastest-growing AI on the Vercel platform in 2026: within the first week of its launch, usage volume surged 27-fold, and the client base expanded by 80 times.
This growing interest in Chinese solutions coincides with tightening AI regulations in the United States. U.S. authorities are considering restrictions on the most powerful models, and OpenAI has already delayed the release of new models at the government’s request. However, similar limitations on advanced AI models may also be imposed in China.
About the Author
