Alibaba’s Qwen family of AI models has crossed 3 billion global downloads, a number that puts it ahead of virtually every open-source competitor in the large language model space.
The milestone cements Qwen’s position as the most downloaded open-source AI model family in the world, with the models now accounting for more than 50% of global open-source model downloads.
Growth that looks more exponential than linear
By January 2026, Qwen had racked up over 700 million downloads on Hugging Face alone, averaging roughly 1.1 million downloads per day during peak periods. By March, cumulative Hugging Face downloads had approached 1 billion.
Developers have been flocking to Qwen for fine-tuning and deployment, building an ecosystem of over 200,000 derivative models on Hugging Face.
When Alibaba launched the Qwen app in public beta back in November 2025, it pulled in 10 million downloads during its first week.
The competition is eating dust
The open-source LLM race has been a three-way fight between Qwen, Meta’s Llama, and DeepSeek. With more than half of all global open-source model downloads now going to Qwen, Alibaba has opened up a meaningful lead.
The latest entry in the pipeline is Qwen3.8-Max, a model with 2.4 trillion parameters that’s slated for an open-weight release in August 2026. It’s designed specifically for coding and agentic tasks.
Why this matters beyond download counts
Over 200,000 derivative models means developers aren’t just downloading Qwen. They’re customizing it, fine-tuning it for specific industries and use cases, and building products on top of it.
Alibaba has been exploring revenue-sharing mechanisms for large-scale commercial users who deploy Qwen through Alibaba Cloud.