The top 10-ranked Chinese LLMs were dominated by Alibaba’s Tongyi Qianwen series, developed in Hangzhou. The Qwen-72B-Instruct LLM topped the list with an average score of 43.02 across six benchmarks. Hugging Face co-founder praised its performance in maths, long-range reasoning, and knowledge. Other Alibaba models, Qwen-72B and Qwen1.5-110B, also made the list. The Yi-1.5-34B-Chat model from Beijing-based start-up 01.AI ranked seventh. Alibaba’s commitment to open-source development has contributed to its rapid progress in AI.
The methodology for ranking AI models by Hugging Face added more metrics, focusing on open-source-developed programs. This excludes advanced LLMs developed in a proprietary environment like OpenAI’s GPT. The University of California, Berkeley, and the Large Model Systems Organisation’s LMSYS Chatbot Arena ranking was topped by OpenAI’s GPT-4o, with Alibaba’s Qwen2-72B-Instruct ranking 19th.
Alibaba’s Tongyi Qianwen models have been adopted by over 90,000 corporate clients, showcasing their widespread use in various industries. The recognition of Alibaba’s LLMs highlights the company’s strong commitment to AI development and open-source technologies, which have been crucial to China’s tech industry growth. Overall, the Chinese LLMs, especially those from Alibaba, are making significant strides in the AI and machine-learning space.
Source link
Source link: https://amp.scmp.com/tech/big-tech/article/3268334/alibabas-large-language-model-tops-global-ranking-ai-developer-platform-hugging-face