AI Photo: VCG

Chinese technology giant Baidu said on Monday it will make its Ernie generative AI large language model (LLM) open-source amid an increasingly fierce AI competition in the world.

By making Ernie's pre-trained weights and inference code fully open-source, Baidu aims to lower access barriers for developers and researchers, fostering a more collaborative open-source AI ecosystem, according to the information the company shared with the Global Times on Monday.

Industry analysts said that Baidu's decision to open-source Ernie marks a pivotal moment in China's AI sector. The move is an important effort by Baidu to compete with global AI players, as well as a significant step to accelerate AI adoption and innovation. 

Open-source allows researchers, developers, and users to access Ernie's underlying code and its "weights" - the parameters that determine how the model processes information - enabling them to use, modify or enhance the model as needed.

"Through open source, Chinese AI models provide developers with an open platform for innovation and application, accelerating AI adoption and benefiting the entire tech ecosystem," Tian Feng, president of Fast Think Institute and former dean of Chinese AI software giant SenseTime's Intelligence Industry Research Institute, told the Global Times on Monday.

Baidu is among a group of Chinese tech companies to invest heavily in AI following the emergence of DeepSeek. As a result, many premium open-source models have emerged. For example, Alibaba Cloud's Qwen2.5-Max has achieved impressive results with over 200 million global downloads and more than 100,000 derivative models, surpassing Meta's Llama as the world's largest open-source model group, China's domestic news outlets reported.

In recent years, China has seen marked advances in open-source LLMs, with numerous AI models gradually catching up with international front-runners in terms of application and performance, Tian said.

According to a CNBC report, Baidu's open-source Ernie is viewed as a "threat" to US OpenAI, Anthropic, and the Chinese rival DeepSeek.

Tian noted that Chinese tech firms are embracing open-source primarily to accelerate AI adoption and innovation, reduce costs, and drive product innovation. This trend is poised to propel China's AI sector forward swiftly and close the technological gap with foreign peers.

Baidu's initiative aligns with China's broader strategy to build an autonomous and globally competitive AI ecosystem, Tian said. According to data from China's Ministry of Industry and Information Technology, China has become the second-largest contributor in open-source AI projects, with the fastest growth rate.

Liu Gang, chief economist at the Chinese Institute of New Generation AI Development Strategies, noted the importance of open-source innovation in driving China's tech advancement. "Through open-source, we can achieve foundational software and hardware integration. This, combined with the deep application of AI in vertical industries, forms the basis of China's AI industry," Liu told the Global Times on Monday.

Despite the optimistic outlook, experts also pinpointed some potential challenges. Liu noted that while China's computational power ranks second globally, the exponential growth in demand for open-source LLMs could create a significant gap in computational resources, possibly hindering the development of China's open-source AI industry.

China has a strong demand for innovative AI products, including DeepSeek, Baidu's Ernie and ByteDance's Doubao. Data indicates that by 2024, nearly 200 generative AI models were registered and went into commercial service in China, with over 600 million registered users, Xinhua News Agency reported earlier.

Global Times