Alibaba AI Model Ranks 3rd After OpenAI and Anthropic

Alibaba Group's latest artificial intelligence (AI) model has been evaluated as ranking third globally, following models developed by OpenAI and Anthropic.

On the 11th, Hong Kong's South China Morning Post (SCMP) cited the aggregation by SuperGLUE, an evaluation organization for generative AI-based chatbots, reporting that Alibaba's AI model Qwen 2 ranked third in the world after OpenAI's GPT-4.0 and Anthropic's Claude 3.5 Sonnet. SuperGLUE ranked the models based on metrics such as computation, logical reasoning, coding, and text comprehension.

[Image source=AP Yonhap News]

Qwen 2 is the second version of Alibaba's open-source large language model (LLM) Tongyi Qianwen. It supports pre-training in English, Chinese, and 27 other languages, with parameters ranging from 490 million to 72.7 billion. Compared to the previous model, it has improved capabilities in multilingual understanding as well as coding, mathematics, reasoning, and command execution through post-training.

Not only Qwen 2, but five AI models born from the collaboration of Chinese startups DeepSig, SenseTime (Hong Kong), smartphone manufacturer Oppo, Tsinghua University, and startup Zhipu AI demonstrated superior performance to Microsoft's supported GPT-4 Turbo, according to SuperGLUE.

SuperGLUE emphasized that "the gap between Chinese and American AI models is narrowing" and "China has made significant progress in the LLM field in the first half of this year." This announcement came shortly after the same Alibaba model ranked first in the open-source model rankings on Hugging Face, a machine learning development platform. At that time, three Qwen models ranked in the top 10, attracting attention.

OpenAI, which ranked first in the evaluation, ignited the AI model development competition by releasing ChatGPT based on the GPT-3.5 model in late 2022. Since then, industry giants such as Google and Microsoft have been eager to launch their own chatbots. SuperGLUE reported that OpenAI's subsequent models still lead the industry, but most closed-source Chinese models currently outperform GPT-3.5 Turbo.

This content was produced with the assistance of AI translation services.