China’s latest wave of open-source artificial-intelligence models, led by DeepSeek’s R1 and Alibaba’s Qwen, is winning tens of millions of users worldwide and prompting fresh concern in Washington and Silicon Valley. DeepSeek’s R1 application expanded from 33 million active users in January to 97 million by April, according to Foreign Affairs, while derivative versions have been downloaded 2.5 million times. U.S. officials and technology executives worry that the low-cost, customizable models could boost Beijing’s soft power in developing markets and erode America’s influence over global AI standards. The surge in Chinese open models comes as OpenAI—long viewed as the pacesetter in generative AI—grapples with a choppy debut of its GPT-5 system. Early adopters and independent benchmarks describe the new model as only marginally better than its predecessor, with some tests showing little gain over rivals from Anthropic and xAI. Company chief executive Sam Altman acknowledged this week that while underlying models are ‘still getting better at a rapid rate,’ chatbots like ChatGPT are ‘not going to get much better.’ OpenAI has moved to mollify dissatisfied users by reinstating its GPT-4o model for subscribers and introducing manual controls that let customers toggle between ‘Auto,’ ‘Fast’ and ‘Thinking’ modes. The firm also raised the weekly cap on complex ‘GPT-5 Thinking’ queries to 3,000 messages. Yet criticism continues, with AI pioneer Geoffrey Hinton joking that GPT-5 could represent a ‘small backwards step’ toward artificial general intelligence. Analysts say the twin developments—China’s rapid progress with open-source models and OpenAI’s rocky upgrade—underscore a shifting competitive landscape. U.S. companies may be forced to open more of their own technology or cut prices to match China’s accessible approach, while policymakers weigh how export controls and alliances can counter the expanding reach of Chinese AI platforms.