Ali AI Reaches Global Open Source Summit, Proving China's Tech Companies Should Not Be Underestimated!

China is making impressive strides in large-scale AI models, with Alibaba's Tongyi Qianwen setting a new standard in the industry. On June 7, Alibaba Cloud launched its latest open-source model, Qwen2-72B, which surpasses the leading U.S. open-source model, Llama3-70B, as well as several closed-source Chinese models, including Wenxin 4.0 and Doubao Pro.

Just two hours after the release, Clément Delangue, co-founder and CEO of Hugging Face, announced that Qwen2-72B had claimed the top spot in the Hugging Face open-source model rankings. This achievement underscores the effectiveness of China's open-source strategy for AI and showcases Alibaba's technological capabilities.

While primarily recognized for its e-commerce ventures, Alibaba's technological expertise is often overlooked. Alibaba Cloud is not only a leader in China's cloud computing sector, but it also stands out as the only global cloud service provider focused on advancing AI models and pursuing an extensive open-source strategy. Over the past year, the Qwen series has consistently ranked among the top Chinese large models, starting with Qwen-14B and followed by the 72B and 110B versions, each achieving improved scores. Notably, Qwen is the only Chinese model featured in an OpenAI-endorsed ranking.

These advancements signal that Tongyi Qianwen is currently the only Chinese model capable of competing with leading international players. In addition to its open-source efforts, Alibaba Cloud is significantly reducing the costs associated with large-scale models by enhancing computing power. Recently, the company announced substantial price cuts for nine Tongyi models, including Qwen-Long, its flagship model akin to GPT-4, which experienced a remarkable 97% price reduction. After the adjustment, Qwen-Long is now available at just 1/400 of GPT-4's cost.