DeepSeek spurs Baidu, other AI competitors to adopt open-source strategy

DeepSeek is spurring competitors at home and abroad to consider an open-source strategy in artificial intelligence (AI) development, following the Chinese start-up’s release of powerful low-cost models that upended how large language models (LLMs) are trained.

Chinese internet search giant Baidu and ChatGPT creator OpenAI are among the nascent field’s major players who have shifted their attention to open-source development, which would enable improved efficiency and a broader adoption of their AI products.

Baidu on Friday said the next generation of its Ernie LLM will be open source from June 30, according to a statement from the Beijing-based company, in a 180-degree turn from founder, chairman and chief executive Robin Li Yanhong’s long-held backing of closed-source AI development.

That followed Baidu’s announcement on Thursday that its Ernie Bot service will be free to use from April 1, providing wider access to the company’s Ernie LLMs. That will end a 17-month-long push to monetise the service, which currently charges users a monthly fee of 49.90 yuan (US$6.84) for access to the firm’s AI models.

Baidu’s major shift in strategy reflects heightened competition in China’s AI market, where Hangzhou-based DeepSeek and internet peer Alibaba Group Holding have recently made big strides in the technology’s development and increased adoption. Alibaba owns the South China Morning Post.