Chinese AI Models Gain Rapid Traction in US Market Despite Trade Restrictions

Chinese artificial intelligence models are seeing a significant increase in adoption by American companies, with some data indicating that over 30 percent of tokens processed by US firms via the OpenRouter platform have been handled by Chinese models since February 8. This trend is driven by the release of high performance models such as Moonshot AI's Kimi K3 and Alibaba's Qwen3.8 Max, which offer competitive performance compared to US leaders like OpenAI and Anthropic at a lower cost. Alibaba's Qwen3.8 Max, featuring 2.4 trillion parameters, is priced at 2 dollars per million input tokens and 6 dollars per million output tokens, which is substantially cheaper than GPT 5.6 Sol and Claude Fable 5. Many of these Chinese models are open source or provide available weights, allowing US companies to host them on local servers to avoid sending data to China. This shift has created internal friction within the US government and tech industry. While some officials suggest sanctions over intellectual property theft, a group of major tech firms including Microsoft, Nvidia, Meta, and Palantir have urged lawmakers not to restrict open models. Meanwhile, OpenAI and Anthropic have raised security concerns regarding Chinese models and reported that their own systems exhibited unpredictable behaviors during cybersecurity tests. The US government continues to attempt to slow this progress by restricting the sale of advanced chips and equipment to China.