Alibaba's Qwen models have surpassed the 3 billion download mark: a new stage in the open AI race.
Chinese tech giant Alibaba has announced an impressive achievement: the total number of downloads of their open-source AI models from the Qwen family has exceeded 3 billion over the past six months. This is a landmark event that is fundamentally reshaping the balance of power in the global artificial intelligence market.
The corporation has provided the developer community with access to more than 460 neural networks, based on which about 300,000 derivative modifications have already been created. These figures demonstrate not just popularity, but the formation of an entire ecosystem around Alibaba's products.
Hugging Face platform data: dominance is obvious
Analytics published by the leading platform for the ML community confirm this trend. Over seven months of the current year, Qwen models were downloaded 2.05 billion times, and the number of their derivative repositories reached 151,448. For comparison, the closest competitor Google has only 82,506. The figures of other giants look even more modest: Google model downloads are estimated at 418 million, and Meta at 227 million. It is important to note that the platform's statistics only account for internal activity, excluding API requests and private deployments, which only underscores the real scale of Qwen's leadership.
The community activity around Qwen is striking: from 180 to 210 new derivative repositories are created daily. Of the 28,531 GGUF conversions of these models, optimized for local execution, only 54 belong to Alibaba itself — the rest have been created by independent developers. This indicates that Qwen has become the de facto standard for fine-tuning and deployment.
The secret of success: scaling strategy and openness
Qwen's success is explained by three key factors. First, regular updates to the model lineup. Second, impressive coverage across all scales: from compact versions with sub-billion parameters to the flagship Qwen3.8-Max with 2.4 trillion parameters. However, the decisive factor was the Apache 2.0 license, which removes all restrictions on commercial use and modification.
Market analysis shows that it was precisely the breadth of the lineup that gave Qwen such an advantage. Models with fewer than 1 billion parameters account for 83% of all downloads on the platform, while giant neural networks with over 100 billion parameters occupy only 1%. Chinese laboratories focused exclusively on large LLMs, such as Moonshot AI, lag significantly behind, having gathered only 37 million downloads in a year.
Geopolitical shift in open AI
The reports demonstrate a fundamental shift: in five of the seven months of the current year, the largest open model from China surpassed all American releases in size. Chinese developers in the segment of models with over 20 billion parameters have almost completely abandoned proprietary licenses: 59% are released under Apache 2.0, 22% under MIT. American companies in this category, meanwhile, use proprietary terms for 41% of models.
It is telling that in the United States, the leaders in the number of new open models were not AI laboratories, but chip manufacturers — AMD and Nvidia, each of which released more than 200 repositories.
My comment: We are witnessing not just competition between technologies, but a clash of two philosophies of AI development. China's open strategy, backed by aggressive scaling, is creating a powerful network effect that, in the long term, may prove more important than control over individual cutting-edge models. This forces a reconsideration of the traditional view of US technological leadership in this field.