Alibaba Group Holding‘s family of large language models (LLMs) – the technology underpinning generative artificial intelligence (AI) services like ChatGPT – has been ranked among the world’s top 10 open-source models.

Three of the four top 10-ranked Chinese LLMs were from the Tongyi Qianwen series, also known as Qwen, developed by e-commerce and cloud computing giant Alibaba, according to AI and machine-learning developer platform Hugging Face, which released its updated leaderboard with new metrics on Wednesday. Alibaba owns the South China Morning Post.

Hangzhou-based Alibaba’s Qwen-72B-Instruct LLM topped the list with an average score of 43.02 across six benchmarks. This Qwen model was trained on 72 billion “parameters” – a machine-learning term for variables present in an AI system during training, which helps establish how data prompts yield the desired output.

Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP Knowledge, our new platform of curated content with explainers, FAQs, analyses and infographics brought to you by our award-winning team.

“Qwen 72B [Instruct] is the king and Chinese open models are dominating overall,” Hugging Face co-founder and chief executive Clement Delangue said in a post on X, formerly Twitter, on Wednesday.

The Tongyi Qianwen large language models, developed and operated by Alibaba Group Holding’s cloud computing unit, are used in industries ranging from consumer electronics to cars and online games. Photo: Shutterstock alt=The Tongyi Qianwen large language models, developed and operated by Alibaba Group Holding’s cloud computing unit, are used in industries ranging from consumer electronics to cars and online games. Photo: Shutterstock>

That particular Qwen model was credited for “its performance in maths, long-range reasoning and knowledge”, according to Wednesday’s post on Hugging Face developer community’s website.

Meanwhile, Alibaba’s Qwen-72B and Qwen1.5-110B models made it to the third and tenth spots, respectively, as of Thursday. The rankings are expected to change over time, as more LLMs get evaluated, according to Hugging Face.

The Yi-1.5-34B-Chat model from Beijing-based start-up 01.AI was ranked seventh. The firm was founded in 2023 by prominent venture capitalist Lee Kai-fu, who was formerly president at Google China.

The lofty recognition of Alibaba’s LLMs shows the company’s rapid progress in AI on the back of its strong commitment to open-source development. Tongyi Qianwen was opened to third-party developers over several months already.

Open source gives public access to a program’s source code, allowing third-party software developers to modify or share its design, fix broken links or scale up its capabilities. Open-source technologies have been a huge contributor to China’s flourishing tech industry over the past few decades.

Alibaba Cloud chief technology officer Zhou Jingren said in May that the Tongyi Qianwen large language model family has been adopted by more than 90,000 corporate clients. Photo: Handout alt=Alibaba Cloud chief technology officer Zhou Jingren said in May that the Tongyi Qianwen large language model family has been adopted by more than 90,000 corporate clients. Photo: Handout>

Facebook parent Meta Platforms‘ AI model took second spot behind Alibaba in Hugging Face’s top 10 list. One LLM from Microsoft, a major backer of ChatGPT creator OpenAI, was ranked sixth.

The Hugging Face community said on Wednesday that its methodology for ranking AI models added more metrics, such as complex problems of around 1,000 words in length, at a time when AI “performances are plateauing”.

Since the Hugging Face ranking is focused on open-source-developed programs, it excludes advanced LLMs developed in a proprietary environment like OpenAI’s GPT.

The LMSYS Chatbot Arena ranking by the University of California, Berkeley, and the Large Model Systems Organisation was topped by GPT-4o, OpenAI’s latest LLM introduced in May. Alibaba’s Qwen2-72B-Instruct was ranked 19th.

This article originally appeared in the South China Morning Post (SCMP), the most authoritative voice reporting on China and Asia for more than a century. For more SCMP stories, please explore the SCMP app or visit the SCMP’s Facebook and Twitter pages. Copyright © 2024 South China Morning Post Publishers Ltd. All rights reserved.

Copyright (c) 2024. South China Morning Post Publishers Ltd. All rights reserved.



Leave a Reply

Your email address will not be published. Required fields are marked *