Alibaba's Qwen Sets a New Record Among Open AI Models

Alibaba Qwen, the open-source large language model (LLM) developed by Chinese tech giant Alibaba, has just set a new world record. According to the latest figures published on January 21 by the AI community Hugging Face, the number of Qwen derivative models has passed 200,000, making it the world's first open-source LLM to reach that mark. The milestone reflects China's growing influence in artificial intelligence and the rising popularity of open-source models.

Proprietary models such as OpenAI's GPT-4 have dominated the headlines, but open models such as Meta's Llama and Alibaba's Qwen are quickly gaining ground by giving developers and researchers more flexibility and control. The Qwen model family has also posted impressive download numbers: more than 1 billion downloads, averaging 1.1 million per day.

That is well ahead of Llama (the US model) and makes Qwen the most downloaded open-source large language model in the world. Download volumes this high point to strong interest in Qwen and broad adoption across a range of applications and research. Open-source models make research more transparent, because anyone can inspect the code and understand how the model works and how it was trained.

This matters especially for tackling bias and fairness issues in AI. Several factors help explain Qwen's success. First, it is a powerful and efficient model trained on a massive amount of data.

Second, Alibaba released it under an open-source license, letting developers and researchers freely use, modify, and distribute the model. Third, Alibaba actively supports the Qwen community with documentation, tools, and assistance. Qwen's rising popularity has significant implications for the AI industry.

It shows that open models are becoming increasingly competitive with proprietary ones, which could drive more innovation and help democratize AI. It could also lower the cost of building and deploying AI, since developers can build on existing open models instead of creating their own from scratch. In short, Qwen's 200,000 derivative models and 1 billion downloads mark an important milestone for the open AI community.

They show the strength and potential of open models, as well as China's growing influence in AI. In the coming years, we can expect even more innovation and competition in AI, driven by the rise of open models like Qwen.

*Meta is designated an extremist organization and banned in the Russian Federation.