Even as the United States and China remain locked in a high-stakes race over artificial intelligence, Chinese-developed AI models are quietly gaining traction in the US market.

Despite geopolitical tensions, a growing number of American programmers and companies are turning to Chinese open-source AI models, which allow users to customise software components.

This contrasts with closed systems such as OpenAI’s ChatGPT or Google’s Gemini, whose internal workings are tightly controlled.

Global adoption of Chinese open-source models surged from 1.2% in late 2024 to nearly 30% by August this year, according to a December report from the developers’ platform OpenRouter and the US venture capital firm Andreessen Horowitz.

Experts say cost is a major factor. “They are cheap – in some cases free – and they work well,” said Wang Wen, dean of the Chongyang Institute for Financial Studies at Renmin University of China. One US entrepreneur told AFP that using Alibaba’s Qwen models instead of proprietary alternatives saves their firm about $400,000 a year.

Chinese models are also being used in parts of the work carried out by US chipmaker Nvidia, AI firm Perplexity, and Stanford University.

The launch of DeepSeek’s open-source “R1” language model in January challenged the perception that top-tier AI must come from US companies such as OpenAI or Google. Other Chinese firms, including MiniMax and Z.ai, have also found overseas users.

Meanwhile, China is advancing in AI “agents” – systems designed to complete online tasks. Open-source models like Moonshot AI’s Kimi K2 are seen as a potential next frontier.

The US government has warned of the strategic importance of open-source models, but many American firms are increasingly focusing on closed systems, even as Chinese alternatives expand globally.