Alibaba’s family of Qwen AI models have topped 700 million downloads on HuggingFace, according to new data, making the Chinese LLM the world’s most widely used open-source AI system.

Fresh research from tech consultancy AIBase, citing figures from open-source platform HuggingFace, shows that Qwen logged more December downloads than the next eight leading AI models combined, easily outpacing the offerings from better‑known competitors like OpenAI, Meta and domestic rival DeepSeek.

The sheer volume of downloads cements Qwen’s status as the most widely used open-source AI ecosystem, building on the momentum of the Qwen App, which reportedly saw more than 10 million downloads within a week of its beta launch in November.

Developed by China’s e-commerce giant Alibaba, the Qwen models run the gamut from the lightweight Qwen3-0.6B to the ultra-large-scale Qwen3-Max, with the AI already used across a variety of consumer-facing products, such as Ant Group’s health app A-Fu, which counts 30 million monthly active users.

“Tens of thousands of real-world applications around the globe have been built based on Qwen, marking a historic peak for [Chinese] open-source large models within the international developer community,” the AIBase report said.

For Alibaba, the success of Qwen reinforces ambitious AI investment plans, with the firm last year committing at least $53 billion (£39.3bn) over the next three years to supercharge its cloud and AI infrastructure, including the development of new foundational models and AI-native applications.

The company has already begun reorganising around that vision, having recently created the Qwen Consumer Business Group, a new unit that plans to consolidate its fast‑growing AI portfolio, including products like the Qwen chatbot, the Quark AI assistant, and the UC Browser, while pushing to integrate lifestyle and productivity services like digital maps, food delivery, office tools, and e-commerce.

However, despite those lofty ambitions, one of China’s most prominent AI researchers recently offered a stark reality check, warning that the country is unlikely to rival US tech giants in developing cutting‑edge models over the next five years.

Recommended reading

• Could DeepSeek Topple the US’s AI Market Dominance?
• Is the UK Lagging Behind in Global AI Race?
• Gartner: Global AI Chips Revenue to Total £56BN in 2024

In comments at an industry conference in Beijing reported by the South China Morning Post, Lin Junyang, a technical lead of the Qwen team, said that there was a less than 20% chance China could match the resources that American tech firms are pouring into their AI development, saying that Chinese compute power is already “stretched to the limit”. 

That view by an insider is decidedly at odds with recent research from the prestigious Human-Centered AI Institute at Stanford University, which said that after years of lagging behind, China’s AI models have “caught up or even pulled ahead” of global counterparts in advanced capabilities. 

The Stanford paper argued that Chinese firms like Alibaba, Tencent, Baidu, Huawei, and ByteDance are leading in computationally efficient models that are beginning to see a larger reach than US peers, a success that may even be helped along by US restrictions on exports of powerful AI chips that reward more efficient open-weight models.

As Chinese AI models become more popular, questions around security become more critical. For example, last year, US government researchers found that DeepSeek was twelve times more susceptible to hijacking attacks than frontier Western models, not to mention the privacy, data, and access issues that arise when using AI created by firms overseen by the Chinese government.