Alibaba’s open-source LLM ranks high in Hugging Face’s top 10 AI models

Open-source artificial intelligence (AI) developer platform Hugging Face has released the latest rankings for large language models (LLMs) using a range of metrics to curate the top 10 models.

According to a report by South China Morning Post (SCMP), the latest ranking focused on open-sourced LLMs while eliminating those created under proprietary conditions. Flowing from the focus on open-source models, mainstream offerings like OpenAI’s ChatGPT and Google’s Gemini (NASDAQ: GOOGL) were noticeably absent from the rankings.

A common theme in Hugging Face’s list is the presence of Chinese LLMs outshining their counterparts from North America and Europe, with Alibaba’s (NASDAQ: BABAF) AI offerings receiving the most plaudits.

The company’s Qwen-72B-Instruct LLM came in first place with a 43.02 score after rigorous testing across multiple benchmarks. Released in 2023, the model, which is part of the Tongyi Qianwen series, closed the gap between it and industry leaders by adopting an open-source model while supporting enterprises in rolling out their own AI products.

“Qwen 72B (Instruct) is the king and Chinese open models are dominating overall,” said Hugging Face CEO Clement Delangue.

The dominance of the smaller Tongyi Qianwen models was captured in the rankings as they occupied fourth and tenth place for their stellar performances in math, logic, and long-range problem-solving skills.

Meta’s Llama (NASDAQ: META) came in second place with a cumulative of 36.67, nearly seven points behind Qwen 72B-Instruct despite its late entrance into the space. A close look at the rankings reveals that Llama 3 has a faltering grasp on second place, with its second and third place within touching distance at 35.59 and 34.35, respectively.

Yi-1.5, another Chinese-based LLM that made the cut, ranked in seventh place, scoring 28.11 across the six metrics, while Microsoft’s (NASDAQ: MSFT) small-language model (SML) Phi-3 joined the exclusive list.

“There are indications that AI builders have started to focus on the main evaluations too much at the expense of model performances on other ones,” Delangue said on LinkedIn. “Bigger is not always smarter.”

The rise and rise of Chinese LLMs

While China lags on the hardware side, AI developers from the country are rolling out impressive offerings to match its peers in other regions. The developers are motivated by the need to tailor AI models to capture the nuances of the Chinese language and prevent foreign products from dominating the local ecosystem.

Alibaba shuttered its quantum computing and emerging technologies arm to focus on AI, a gamble that has paid off handsomely for the company. As OpenAI prepares to limit access to its API in China and other restricted areas, several Chinese firms are angling to provide similar services to AI developers in a frantic race to increase their market shares. 

In order for artificial intelligence (AI) to work right within the law and thrive in the face of growing challenges, it needs to integrate an enterprise blockchain system that ensures data input quality and ownership—allowing it to keep data safe while also guaranteeing the immutability of data. Check out CoinGeek’s coverage on this emerging tech to learn more why Enterprise blockchain will be the backbone of AI.

Watch: Calvin Ayre is all in on Metanet—the game-changing fusion of enterprise blockchain, AI & IPv6