Alibaba Qwen AI Models Surpass 3 Billion Downloads, Overtaking Meta and Google

Alibaba's Qwen family of open-source artificial intelligence models has surpassed 3 billion cumulative downloads, according to figures released by the company, placing it ahead of comparable open-source model offerings from Meta and Google. The milestone marks a significant shift in the competitive landscape for publicly available AI models.

What Happened

Alibaba announced that its Qwen series of AI models has crossed the 3 billion download threshold, a figure the company says exceeds the download counts recorded by open-source AI models released by Meta and Google. The announcement was reported by entARABI and positions Qwen as the most widely downloaded open-source large language model family currently available.

Alibaba has not specified the precise timeframe over which the 3 billion downloads accumulated, nor has it detailed the breakdown of downloads across individual model versions within the Qwen series.

Background

Alibaba's Qwen models are developed by the company's cloud and AI division, Alibaba Cloud. The Qwen series includes multiple model sizes and variants, ranging from smaller models suitable for edge deployment to larger models targeting enterprise and research applications. Alibaba has released successive generations of Qwen models in recent months, each accompanied by benchmark performance claims across standard AI evaluation tasks.

Meta's Llama series and Google's Gemma models are among the most prominent competing open-source model families. Meta in particular has positioned the Llama series as a flagship open-source offering, reporting hundreds of millions of downloads since the original Llama release in 2023. Google's Gemma models, released in early 2024, have also accumulated substantial download figures across developer platforms including Hugging Face.

Open-source model downloads are typically tracked through repositories such as Hugging Face, GitHub, and proprietary cloud marketplaces. Alibaba has not publicly specified which platforms were counted in its 3 billion figure.

What the Numbers Indicate

The 3 billion download figure, if verified across standardised counting methodologies, would represent the largest reported cumulative download total for any single open-source AI model family to date. Download counts in the AI model space are a commonly cited proxy for developer adoption and ecosystem reach, though they do not directly correspond to active deployment or commercial usage.

Alibaba Cloud has been expanding the Qwen model family's international presence, offering access through its ModelScope platform as well as third-party repositories. The company has increasingly targeted developers and enterprises outside mainland China as part of its cloud expansion strategy.

Meta reported in April 2024 that Llama models had been downloaded more than 350 million times. Alibaba has not published a direct side-by-side comparison methodology against Meta or Google figures, and the basis for its claim of surpassing both companies has not been independently verified at the time of this report.

Context in the Broader AI Market

The open-source AI model market has grown rapidly over the past 18 months, with Chinese technology companies including Alibaba, Baidu, and DeepSeek increasing their presence alongside established Western players. DeepSeek's R1 model release earlier in 2025 drew significant international attention after benchmark results suggested competitive performance with models from OpenAI and Anthropic.

Alibaba's download milestone arrives as the company continues to invest in AI infrastructure across its cloud division, which has reported AI-related revenue growth in consecutive quarters. The Qwen models serve as a foundation for enterprise customers building customised AI applications on Alibaba Cloud's platforms.

Alibaba has not announced a specific date for the next major Qwen model release, though the company has indicated continued development of the series is ongoing.