Official Statement: Qwen Models Will Stick to Open Source to the End
Qwen leadership officially confirmed that future Qwen 4.0 and 5.0 models will remain open source, dispelling rumors of a shift to a closed API-only model. Liu Dayi Heng, head of Qwen, stated that the project aims to make AI a foundational capability accessible to users globally. This announcement ensures the continued availability of top-tier open-source LLMs for developers and researchers worldwide, preventing fragmentation in the open AI community. It reinforces Alibaba's commitment to democratizing artificial intelligence access against industry trends toward proprietary systems. The Qwen series has achieved over 3 billion cumulative global downloads, with more than 460 open-source models and over 300,000 derivative models created. This milestone places the Qwen family at the forefront of the global open-weights AI ecosystem.
## BACKGROUND
Qwen is a prominent family of frontier multimodal AI models developed by Alibaba's Tongyi Lab. Open-source or open-weight AI models allow developers to freely download, run, and fine-tune model parameters for various custom applications.