01.AI

What it is

01.AI is a Chinese artificial intelligence company founded in Beijing in March 2023 by Kai-Fu Lee, a former Microsoft and Google executive who co-founded the venture firm Sinovation Ventures. Lee serves as its chief executive ^[1]. Operations began in June 2023 ^[1]. The company's Chinese name, 零一万物, alludes to the Tao Te Ching, a classic Chinese philosophical text ^[1].

Origins and funding

Among the company's backers are Alibaba Group and Xiaomi. It also borrowed money from Sinovation Ventures to buy semiconductors, the chips used to train and run AI models ^[1].

Within about eight months of its founding, 01.AI reached a valuation of more than US$1 billion. The valuation followed a funding round that included Alibaba's cloud unit, and by November 2023 the company had become a unicorn, the industry term for a start-up valued at more than US$1 billion ^[2].

Models and products

The company's first models were the open Yi-6B and Yi-34B base models, released on 2 November 2023. Base models are trained to continue and understand text in general, and they are often adapted later into chat assistants. At the time, Yi-34B ranked first among pre-trained open models on the Hugging Face Open LLM Leaderboard and on C-Eval, a benchmark that tests knowledge and reasoning in Chinese ^[3].

On 13 May 2024, at its first-anniversary event, 01.AI launched Yi-Large, a closed model with about one hundred billion parameters, which means users reach it through the company's service rather than downloading its weights. The same event open-sourced the Yi-1.5 series, opened the company's API platform to users worldwide and launched Wanzhi, a free consumer productivity app ^[4]. An API platform is the interface that lets other developers send requests to a model from their own software.

The company kept releasing models through 2024. It released Yi-Coder, a family of code models, in September 2024, and Yi-Lightning, its flagship model built on an enhanced Mixture-of-Experts design, in October 2024. A Mixture-of-Experts model splits its network into many specialised parts and activates only some of them for each input, which saves computing power. In the LMSYS ranking of model makers, 01.AI tied with xAI for third place, behind OpenAI and Google ^[1].

On 17 March 2025 01.AI launched the Wanzhi enterprise one-stop LLM platform. It offered enterprise deployment of DeepSeek models and also supported Qwen and the company's own Yi models ^[8].

A change of direction

In early January 2025 reports said 01.AI was selling its pre-training team to Alibaba Cloud. Lee denied the reports. He said Alibaba would take on the training of very large models through a joint lab, while 01.AI's own teams would focus on smaller, cost-effective models. In his view, only tech giants can bear the cost of training super-large models ^[5][6].

Lee also said that business-to-business (ToB) revenue had exceeded 100 million yuan in the previous year, and that the company would not build trillion-parameter models, focusing on lightweight models instead ^[7]. Pre-training is the first and most expensive stage of building a model, when it learns from a large text corpus.

Wikipedia reports that by March 2025 01.AI had stopped pre-training large language models altogether, and was instead focusing on tailored business solutions built on DeepSeek's models ^[1].

Recent developments

On 7 July 2026 01.AI launched a product matrix of AI agents called 一号位AI, aimed at chief executives, heads of sales and heads of investment. Lee said 01.AI was benchmarking itself against Palantir and using forward-deployed engineers, who work at client sites to adapt software to a customer's needs ^[9].

Its place in the field

01.AI's history falls into three phases. In 2023 and 2024 it released open base models and then a closed flagship, and it used public leaderboards such as the Hugging Face Open LLM Leaderboard and the LMSYS Chatbot Arena to show how its models compared with others. In 2025 it narrowed its focus toward smaller models and enterprise deployments, a path that reflects the view that very large models are only affordable for the largest technology companies. The company's shift is one example of how Chinese start-ups have responded to the cost of training frontier models.