Alibaba's Qwen models have surpassed the 3 billion download mark: China strengthens its leadership in open AI.
Chinese corporation Alibaba has reached an impressive milestone: the total number of downloads of its open AI models from the Qwen family has exceeded 3 billion in just the last six months. This is a landmark event that vividly demonstrates the shifting balance of power in the global artificial intelligence market.
The company has made more than 460 different neural networks publicly available. Based on them, the global developer community has created about 300,000 derivative models, indicating the formation of a powerful ecosystem around Qwen.
Hugging Face Data: Impressive Numbers
My analysis of data from the Hugging Face platform over the past seven months confirms this trend. Within this ecosystem alone, Qwen has been downloaded 2.05 billion times, and the number of derivative repositories has reached 151,448. For comparison, Google's figure stands at only 82,506. Over the same period, Google's model downloads totaled 418 million, while Meta's reached just 227 million.
It is important to emphasize that Hugging Face statistics only account for activity within its own platform, excluding API requests and corporate deployments. Thus, the real scale of Qwen's distribution is likely even higher. The number of Qwen derivative repositories grows by 180-210 daily. Notably, out of 28,531 GGUF conversions (a format for local deployment), only 54 were created by Alibaba itself—the rest were done by the community.
The Secret to Success: Strategy, Not Coincidence
I attribute Qwen's superiority over competitors to three key factors. First, regular updates to the model lineup. Second, coverage of all scales—from compact versions with sub-billion parameters to the flagship Qwen3.8-Max with 2.4 trillion parameters. Third, and most importantly, the Apache 2.0 license, which removes all restrictions on modification and commercial use.
It is precisely the breadth of the lineup that has become the decisive factor. According to Hugging Face data, models with fewer than 1 billion parameters account for 83% of all downloads in the platform's history. Meanwhile, neural networks with more than 100 billion parameters account for only 1%. Therefore, laboratories focused exclusively on giant LLMs, such as Moonshot AI, are losing out: their models have been downloaded only 37 million times over the year—55 times less than Qwen. In the local deployment segment (GGUF), Qwen also leads with 39.6 million downloads per month, compared to 20.8 million for Gemma and 7.5 million for Llama.
A Shift in the Balance of Power in Open AI
The Hugging Face report records a fundamental shift. Almost every month this year, the largest open model from China has surpassed all American releases in size. The ceiling for Chinese models ranged from 754 billion to 2.78 trillion parameters, while for US competitors it did not exceed 130 billion in five out of seven months.
The difference in licensing policies is also notable. In China, 59% of models with more than 20 billion parameters are released under Apache 2.0, 22% under MIT, and none have restrictions on commercial use. For American developers, the picture is different: only 29% under open licenses, 41% under proprietary terms, and 30% with no license specified at all.
Interestingly, in the US, the leaders in the number of new open models were not AI laboratories but chip manufacturers: AMD and Nvidia each released more than 200 repositories, leaving Google and Meta far behind. This confirms my long-held view: the US advantage in the AI race currently rests not on algorithms but on control over computing infrastructure.
My conclusion: Alibaba's strategy of democratizing access to AI through open licenses and a broad model lineup has proven not just successful but revolutionary. Chinese models are becoming the de facto standard for developers worldwide, and this will change the competitive landscape of the industry for a long time to come.