Alibaba CEO: Qwen will remain open-source

Alibaba Group CEO Eddie Wu has made a definitive public commitment that the company's Qwen series of large language models will remain open-source. This announcement comes during a period of increasing industry debate about open versus closed AI development, with major players like OpenAI, Google, and Anthropic keeping their most advanced models proprietary. Wu's statement directly addresses concerns from the developer community about potential policy changes, reinforcing Alibaba's strategy to build ecosystem adoption through accessibility rather than through restrictive licensing. The commitment applies to the full Qwen family, including the recent Qwen2.5 series which has shown competitive performance on key benchmarks.

Technically, the open-source pledge covers models like the flagship Qwen2.5-72B, a 72-billion parameter model that rivals proprietary offerings in reasoning and coding tasks. Developers and enterprises can freely use, modify, and deploy these models commercially without paying licensing fees to Alibaba. This open approach contrasts sharply with the 'open-weight' but restricted-use models from Meta (Llama) and contrasts with the fully closed models from other tech giants. The strategy aims to make Qwen the foundation model of choice in China and globally for applications requiring transparency, customization, and cost control, potentially accelerating AI integration in various industries from cloud services to edge devices.