The Beijing-based artificial intelligence startup Baichuan Intelligent Technology announced today that it has raised more than $300 million in a Series A1 round led by major Chinese technology giants Alibaba Group Holding Ltd., Tencent Holdings Ltd. and smartphone maker Xiaomi Corp.

The new funding comes in addition to a previous $50 million angel round, providing Baichuan with $350 million in total. Since being founded in April 2023 and its two funding rounds, with this funding the company has already exceeded $1 billion in valuation, catapulting rapidly into the coveted “unicorn” status.

Baichuan stands out as one of the foremost AI developers in China working on generative AI and large language models and competing with Microsoft Corp. and OpenAI LP. Since its launch, the company has already rapidly rolled out numerous LLMs, including four open-source models that have been downloaded more than 6 million times, and two proprietary models known as Baichuan-53B and Baichuan2-53B.

The company became one of the first companies to receive regulatory approval to release large language models in China and began to roll out its Baichuan2-53B commercially. It received approval alongside five other Chinese tech firms, including Baidu Inc., China’s leading online search provider and maker of a ChatGPT-style chatbot, Ernie Bot, that is now fully accessible to the public, and SenseTime Group Inc., the developer of another chatbot. Just this week, Baidu rolled out Ernie 4.0, the newest version of its model that the company says rivals OpenAI LP’s GPT-4.

China has sought catch up with the U.S. in the AI market, especially with companies such as OpenAI, Anthropic, Cohere and others rapidly developing powerful large language models and art generating models. Part of what has made the competition more difficult is that many of the advanced chips required to train and provide the runtime for powerful AI models are restricted by the U.S. from export to China, meaning that they must be fabricated in country or imported from other countries.

The country has also been working to set stricter rules regarding the production and use of generative AI, especially chatbots available to the general public. Although China is working to encourage the innovative use of AI, it’s also supporting the secure development of AI, and as such it has been fining companies that use it irresponsibly. Most recently China’s National Information Security Standardization Committee instituted a blacklist of training data that cannot be used to train generative AI models, including anything considered that might damage the country’s image, advocate terrorism or overthrow the social system.

These increased regulations and the dearth of advanced chips have led smaller AI startups to find it difficult to gain a foothold in the AI market in China when attempting to develop and train their own AI models.

Baichuan has positioned itself as the developer of LLMs and provides its technologies as a model-as-a-service by provisioning access through application programming interfaces to companies. As a result, Chinese startups looking to create chatbots using text interfaces can use its models by fine-tuning its models at large scale by passing data through its systems and receiving outputs.

Image: Pixabay