AsiaTechDaily – Asia's Leading Tech and Startup Media Platform
Chinese startup 01.AI has rapidly emerged as a unicorn in the tech world, achieving this status in less than eight months with its performance in a new open-source AI model, Yi-34B. The Beijing-based company recently surpassed a $1 billion valuation following a successful funding round.
Led by renowned computer scientist Kai-Fu Lee, 01.AI focuses on developing AI systems specifically tailored to the Chinese market. This places the startup in competition with other domestic companies striving to establish China‘s equivalents to models like OpenAI’s ChatGPT.
According to Lee, the driving force behind 01.AI’s innovation was necessary due to limited access to AI systems provided by tech giants such as Google and OpenAI. He emphasized that many in the large language model (LLM) field are working to create solutions for a market with a strong demand for AI applications.
In a recent funding round that included Alibaba Group Holding’s cloud unit, 01.AI achieved a valuation of over $1 billion. The company began assembling its team in March and commenced operations in June, rapidly rising to unicorn status.
Large language models are advanced algorithms trained on vast datasets to read, comprehend, and generate human-like text, images, and code. 01.AI
The Yi-34B model developed by 01.AI represents a bilingual English/Chinese model created in-house by the company’s researchers. It has a parameter count of 34 billion and is trained on a dataset with sequences as long as 4,000 tokens, expandable to 32,000 during inference.
The flagship product is now available for developers worldwide in both Chinese and English.
Yi-34B outperforms established open-source models, including Meta Platforms’ esteemed Llama 2, on critical metrics. Hugging Face, a platform that maintains leaderboards for evaluating the best-performing Large Language Models (LLMs) across various categories, recently ranked the Chinese model as the top pre-trained base LLM in performance.
The company has adopted a model licensing approach that allows free academic and commercial usage, subject to permissions as outlined in Model License Agreement 2.0, making it an attractive choice for the open-source community and industry users.
Amidst the fierce competition in the AI sector, 01.AI’s strategy of securing semiconductors is crucial in safeguarding its operations against the intricate challenges arising from the U.S.-China tensions that affect access to advanced AI hardware.
By taking early measures, including procuring semiconductors and even acquiring funds to support this effort, the company has ensured a buffer against potential supply chain disruptions, allowing it to maintain operational stability for more than a year.
01.AI has plans to develop models with a parameter count ranging from 100 billion to 400 billion over the next 1-2 years. These forthcoming models are expected to bring advancements beyond the already impressive capabilities of Yi-34B.
Also Read: