Chinese open-source AI models become global standards
Updated
Updated · Foreign Policy · May 7
Chinese open-source AI models become global standards
13 articles · Updated · Foreign Policy · May 7
Kimi K2.6 recently topped OpenRouter usage, while Alibaba's Qwen took more than 50% of global open-source downloads and reached about one billion downloads.
The report says low prices and open-source access are helping Chinese models spread across emerging markets, where firms and governments want cheaper, locally adaptable alternatives to U.S. systems.
It argues Beijing could gain long-term influence by embedding Chinese technical standards globally, echoing Belt and Road logic as countries build on Qwen, Kimi and other Chinese architectures.
What are the hidden costs for developing nations adopting China's AI as their digital backbone?
Is China's 'good enough' AI strategy set to outmaneuver the West's focus on technological supremacy?
Will China's manufacturing dominance create an unbeatable real-world data advantage for its AI models?
Alibaba’s Qwen Hits 1 Billion Downloads, Capturing Over 50% of Global Open-Source AI Market in 2026
Overview
By March 2026, Alibaba's Qwen surpassed 1 billion downloads, capturing over half of all global open-source AI model downloads. This success followed the release of Qwen 2.5 in 2024 and DeepSeek's cost-efficient model, which sparked a price war and accelerated adoption. China's strategic open-source approach, driven by Alibaba's public release of model weights, fueled rapid growth across 119 languages and emerging markets, supported by massive investments in AI infrastructure and talent. However, Qwen's dominance raised Western security concerns, leading to export controls and regulatory challenges that created a hardware bottleneck for China. Despite this, China's open-source strategy is reshaping the AI landscape toward a multipolar future centered on accessibility, customization, and global collaboration.