By Yelin Mo, Josh Ye and Kane Wu
BEIJING/HONG KONG (Reuters) -Beijing-based AI startup 01.AI is raising up to $200 million in fresh capital, two sources familiar with the matter said, building on a valuation of $1 billion reached last month on the back of surging global interest in open-source AI models.
It is one of several Chinese startups that opened, or plan to open, to public use their large language models (LLM), alongside larger firms Meta and Alibaba in a scramble to gain users and catch up with market leader OpenAI.
One of the sources said 01.AI, officially launched by Lee Kai-fu, Google China’s former chief, in July after a three-month incubation period, hit a valuation of $1 billion early in November.
The other source said the company was seeking additional investments from U.S. dollar-based investors.
All the sources cited in this report sought anonymity, as the fundraising details were not public.
The company did not immediately respond to a request for comment.
01.AI drew recognition in the open-source LLM community in November, when its Yi-34B model became the first Chinese LLM to top the leaderboard of Huggingface, a platform for tech firms to share LLMs, which then get rated for performance and popularity.
Riding the interest in generative artificial intelligence (AI) has helped such startups raise significant funds from investors after promising to make their AI models open.
For example, the valuation of Zhipu AI, founded in 2019 has crossed $1 billion, Chinese media have said.
Eight-month-old Baichuan Technology attained a valuation of $1.2 billion after its last fundraising in October, said a third source who had direct knowledge of the exercise.
The movement to open-source LLMs has been criticised by OpenAI, maker of hit chatbot ChatGPT, which has kept the codes of its models under tight wraps, invoking the risk of abuse by malicious actors that could bring danger to society.
In China, tech giant Alibaba’s cloud arm has been active in making its LLMs open-source. Its latest LLM Qwen-72B recently topped Huggingface’s leaderboard, becoming the second model from China to do so.
Qwen-72B is the latest, and most powerful, of eight AI models Alibaba has made open in the last four months. The company says Qwen-72B can outperform OpenAI’s flagship GPT4 in handling Chinese, going by some industry benchmarks.
“Building up an open-source ecosystem is critical to promoting the development of LLM and AI applications building,” Jingren Zhou, the chief technology officer of Alibaba Cloud, told Reuters on Wednesday when asked about Alibaba’s commitment to open-sourcing its AI models.
Hunger for AI applications is growing across a wide range of industries, developers and businesses, he added.
“Alibaba Cloud aspires to become the most open cloud and make generative AI capabilities accessible to everyone.”
The company is also supporting smaller open-source AI LLM startups, including 01.AI, with which it is collaborating on aspects such as model training and deployment processes.
AliCloud earlier committed funding in the current round, the sources said.
AliCloud did not immediately respond to a request for comment on Thursday on the 01.AI fundraising.
(Reporting by Yelin Mo in Beijing, Josh Ye and Kane Wu in Hong Kong; Editing by Brenda Goh and Clarence Fernandez)