Alibaba's Qwen models have surpassed the 3 billion download mark: a new stage in the open AI race
Chinese tech giant Alibaba has reached an impressive milestone: over the past six months, the total number of downloads of its open-source AI models from the Qwen family has exceeded 3 billion. This is a landmark event that is fundamentally reshaping the balance of power in the global artificial intelligence market.
The corporation has provided the community with access to more than 460 neural networks, on the basis of which developers around the world have created about 300,000 derivative models. This scale of ecosystem has no analogues among Western competitors.
Hugging Face Data: Numbers That Speak for Themselves
An analysis of the Hugging Face platform, the largest hub for the AI community, confirms Qwen's dominance. Over seven months of the current year, Alibaba's models were downloaded 2.05 billion times, and the number of forks (derivative repositories) reached 151,448. For comparison, Google's figure is only 82,506.
At the same time, it is important to understand the context: Hugging Face statistics only account for activity within its own ecosystem, excluding API requests and private corporate deployments. Nevertheless, even this incomplete data demonstrates a colossal gap. Over 2026, Google models were downloaded 418 million times, and Meta — 227 million times.
The daily growth is especially telling: the number of Qwen derivative repositories increases by 180–210 per day. The community is actively involved in adapting the models: out of 28,531 GGUF conversions (optimized for local deployment), only 54 were created by Alibaba itself; the rest is the work of enthusiasts. As experts rightly note, Qwen has become the standard workflow for developers choosing a model for fine-tuning and deployment.
The Secret to Success: Scaling Strategy and Openness
Qwen's superiority is explained by three key factors: regular updates, coverage of all scales — from compact versions with sub-billion parameters to the giant Qwen3.8-Max (2.4 trillion parameters) — and the Apache 2.0 license, which does not restrict commercial use.
The breadth of the model lineup proved decisive. According to Hugging Face data, 83% of all downloads in the platform's history come from models with fewer than 1 billion parameters, while giant neural networks (over 100 billion) account for only 1%. Competitors focused on large LLMs are losing out: for example, Moonshot AI, which ignores small models, has gathered only 37 million downloads in a year — 55 times fewer than Qwen.
In the local deployment segment, Alibaba also leads: Qwen GGUF builds are downloaded 39.6 million times per month, outpacing Gemma (20.8 million) and Llama (7.5 million).
A Global Shift in Open AI
The report reveals fundamental changes. Almost every month, the largest open model from China surpassed all American releases in size, reaching a ceiling of 754 billion to 2.78 trillion parameters, while in the US, in five out of seven months, this figure did not exceed 130 billion. Moreover, in China, 59% of models with over 20 billion parameters are released under Apache 2.0, and 22% under MIT, without any restrictions. In the US, 41% of such models have proprietary licenses, and 30% have no license specified at all.
Notably, in the US, the leaders in the number of new open models were not AI labs but chip manufacturers — AMD and Nvidia, each releasing more than 200 repositories. Google and Meta lag noticeably behind.
My analysis: Qwen's success is not just statistics but a signal of a paradigm shift. The Chinese ecosystem is betting on total openness and accessibility, which stimulates mass adoption and creates a powerful network effect. The US, relying on proprietary developments and control over infrastructure, risks losing the initiative in the battle for developers' minds, which ultimately determines technological leadership.