Alibaba Cloud's Qwen Series Accounts for Over Half of Global Open-Source Model Downloads, Totaling Nearly 1 Billion

DeepTech TechFlow reports that, as of April 10, according to the South China Morning Post, a report released by U.S. AI tracking agency Interconnects AI shows that by March 2026, Alibaba Cloud's Qwen series models have accounted for over 50% of global open-source model downloads, with cumulative downloads reaching 942.1 million, far surpassing competitors like Meta's Llama and DeepSeek. In February alone, Qwen's downloads reached 153.6 million, exceeding the combined downloads of the next eight major competitors. The report highlights that Qwen's dominance stems from the widespread adoption of its small-parameter versions (under 10 billion parameters), which allow developers to customize and deploy at low cost. Since the release of Qwen 2.5 in September 2024, Chinese models began to surpass mainstream U.S. open-source models like Llama, and the launch of Qwen 3.5 in February this year further solidified its leading position. Meanwhile, open-source strategies are becoming a critical battleground in U.S.-China AI competition. Meta has abandoned the open-source approach this year, opting instead to release its closed-source flagship model Muse Spark. Similarly, Chinese companies like Alibaba Cloud and Zhipu AI have also transitioned some of their latest models to closed-source to expand direct commercialization channels.