Alibaba's open-source Qwen models have surpassed the 3 billion download mark: a phenomenon that is reshaping the AI market
Chinese tech giant Alibaba has reached an impressive milestone: over the past six months, the total number of downloads of their open-source AI models from the Qwen family has exceeded 3 billion. This is not just a number, but a marker of a tectonic shift in the industry, where open weights are becoming the main battlefield for developers' minds.
The corporation has made more than 460 neural networks publicly available, which have served as the foundation for creating 300,000 derivative models. This indicates the formation of an entire ecosystem around Qwen, rather than just one-off popularity.
Hugging Face Data: Dominance Without Reservation
Independent statistics from the Hugging Face platform paint an even more striking picture. Over seven months of the current year, Qwen models were downloaded 2.05 billion times, and the number of derivative repositories reached 151,448. For comparison, the closest competitor Google has only 82,506. Google's model downloads for 2026 are estimated at 418 million, and Meta's at 227 million.
It is important to understand that these figures only account for activity within the Hugging Face ecosystem, excluding API requests and private deployments. However, even with this caveat, the gap is colossal. Notably, the number of Qwen derivative repositories grows by 180-210 daily, and out of 28,531 GGUF conversions of Alibaba models, only 54 were created by the company itself—the rest is community work. Qwen has become the standard working tool for those deciding which model to fine-tune and deploy.
The Secret to Success: Strategy, Not Chance
Qwen's position is explained by three key factors. First, regular updates to the model lineup. Second, coverage of all scales—from sub-billion versions to the giant 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.
The breadth of the lineup proved decisive. The data shows that 83% of all downloads in the platform's history come from models with fewer than 1 billion parameters, while neural networks over 100 billion account for only 1%. Laboratories focused on giant LLMs, such as Moonshot AI, are losing out: their annual downloads are 55 times fewer than Qwen's. In the local deployment segment, Alibaba's GGUF builds are downloaded 39.6 million times per month, compared to 20.8 million for Gemma and 7.5 million for Llama.
The Geopolitics of Open Source
The report also revealed a strategic shift in the balance. Almost every month, the largest open-source model from China surpassed all American releases in size, reaching a ceiling of 2.78 trillion parameters. Meanwhile, in China, 59% of models with more than 20 billion parameters are released under Apache 2.0, while in the US only 29% have open licenses, with 41% being proprietary. Notably, in the US, the leaders in the number of new open models were not AI labs but chip manufacturers—AMD and Nvidia.
My view: Qwen's success is not just a victory for one company, but a demonstration that an open model with the right license and coverage of all niches becomes an infrastructure standard. While American giants spend resources on proprietary developments, China is methodically capturing the global developer base, laying the foundation for dominance in the next cycle of AI adoption.