Alibaba's Qwen passes 3 billion downloads — a record

Alibaba's open Qwen model family has passed three billion downloads within six months, according to Hugging Face's latest “state of open models” report. That puts Qwen far ahead of the open models of Google (418 million downloads in 2026) and Meta (227 million) — combined.

The ecosystem behind it is vast: Alibaba has open-sourced more than 460 models, from which the community has built over 300,000 derivatives — fine-tuned variants for specific languages, industries, and tasks. According to the report, Qwen has become “part of the default workflow” when developers decide which models to fine-tune and deploy.

Strategically, Alibaba benefits from its cloud distribution: the company delivers Qwen through its cloud platform directly to enterprise customers across Southeast Asia and Africa — reach that many competitors lack. Qwen is also extending its lead over Chinese rivals such as DeepSeek, Moonshot AI, and MiniMax.

The numbers underline a shift in the global AI race: while US providers keep their frontier models largely closed, Chinese labs dominate the open scene — with models that increasingly replicate frontier performance while staying affordable and customizable.

For European companies, the signal cuts both ways: the cost of capable, self-hosted AI is falling dramatically — but dependence is growing on models whose training data and governance sit outside European control.