Alibaba Cloud Qwen series models account for over 50% of global downloads of open-source models, with nearly 1 billion cumulative downloads

Gate News: On April 10, the U.S. AI tracking organization Interconnects AI released a report showing that, as of March 2026, Alibaba Cloud’s Qwen series models have accounted for more than 50% of global open-source model downloads, with cumulative downloads reaching 942.1 million times—far exceeding competitors such as Meta Llama and DeepSeek. In just February alone, Qwen’s download volume reached 153.6 million, surpassing the combined downloads of the eight major vendors ranked after it. The report says that Qwen’s leading position has been driven by the extremely wide adoption of its small-parameter versions (below 100 billion parameters), allowing developers to freely customize and deploy them at low cost. Since the release of Qwen 2.5 in September 2024, Chinese models have begun to outpace mainstream U.S. open-source models such as Llama; after the release of Qwen 3.5 in February 2026, the leading advantage has been further solidified. At the same time, open-source strategy is becoming a key battleground in AI competition between China and the United States. Meta has already abandoned its open-source route this year and instead released a closed-source flagship model, Muse Spark; Chinese companies such as Alibaba Cloud and Zhipu AI have also converted some of their latest models to closed source to expand direct commercialization channels.