Qwen (Tongyi Qianwen) is Alibaba Cloud's AI model family. Since late 2025, Qwen has surpassed Meta Llama as the world's most downloaded open-source LLM, with 600+ million downloads and 180,000+ derivative models.
What is Qwen?
Qwen (short for Tongyi Qianwen, "thousand questions") is the large language model family developed by Alibaba Cloud. Since October 2025, Qwen has surpassed Meta Llama as the world's most downloaded open-source LLM.
Alibaba has published nearly 400 open-source models, and the community has created 180,000+ derivative models based on the Qwen architecture.
Qwen Models
- Qwen 3.5 (February 2026) — 397B parameters, natively multimodal, 201 languages, agentic capabilities
- Qwen 3 — 6 dense models + 2 MoE, from 0.6B to 235B parameters, trained on 36 trillion tokens
- Qwen 2.5-Omni — Multimodal input (text, images, video, audio) with text + audio output
Qwen by the Numbers
- 600+ million downloads for the Qwen 3 family
- 180,000+ community-created derivative models
- 30 million monthly app users (reached in 2 weeks)
- 90,000+ enterprises deployed via Alibaba Cloud & DingTalk
- In December 2025, Qwen's monthly downloads surpassed the next 8 models combined
Qwen and AI Visibility
Qwen is particularly important for brands targeting Asian and Chinese markets. But its open-source dominance means Qwen-based models power many third-party applications worldwide. A comprehensive GEO audit must test Qwen models. AI Labs Audit includes Alibaba/Qwen models in its a wide selection of models. Check our tutorial.