Alibaba Cloud has introduced an updated version of its large-language model called Qwen2.5, which in some ways surpasses OpenAI's popular GPT-4 model.
Alibaba Cloud CTO Jingren Zhou noted that Qwen2.5 shows improvements in the areas of reasoning, code understanding, and natural language compared to the previous version of Qwen2.0.
An analysis of the OpenCompass platform in March mainly shows that Qwen2.5 is superior to GPT-4 in terms of language and creativity. At the same time, the Alibaba model lags behind GPT-4 in aspects such as breadth of knowledge, logical thinking, and mathematical ability.
Large language models like Gwen are trained on large amounts of textual data and used to generate natural language responses. Since the release of ChatGPT in 2022, which caused a buzz around the world, Alibaba has been actively developing AI technologies. In April 2023, the Tongyi Qianwen model was introduced, and in October of the same year, an updated version of the Qwen with expanded functionality was released.
Alibaba Cloud says that Qwen-based services, including enterprise messaging app DingTalk, are already used by more than 2.2 million enterprise customers. In addition, the company offered a number of open Gwen models to the developer community and improved the Model Studio platform with new tools for creating AI.
