Baidu, the leading technology firm in China, is set to open-source its Ernie artificial intelligence (AI) model on Monday. This announcement, originally made in February, follows the company’s decision to offer its chatbot for free. The move is being hailed as a significant development in China’s AI industry, marking the first instance where a prominent AI company has opted to share its proprietary technology with the broader community. Experts suggest that Baidu’s initiative could parallel the influence that DeepSeek-R1 had on the global AI landscape upon its release.
Baidu to Gradually Open-Source Ernie
On February 14, Baidu confirmed that its flagship large language model (LLM), Ernie, would be accessible to the open community. A post on WeChat indicated that the company plans to introduce its “next generation of AI model,” anticipated to be Ernie 5. In March, Baidu released the Ernie 4.5 foundation model and its inaugural reasoning model, referred to as Ernie X1.
As Baidu transitions to an open-source strategy, it remains unclear whether the AI models will be fully open-sourced or only partially. Fully open-sourced models provide developers with model weights, training guidelines, and architectural details, allowing them to use, modify, replicate, and build upon the models freely.
Partial open-sourcing, on the other hand, involves making only the model weights available, while keeping other crucial details under wraps. In such cases, developers can use and fine-tune the model but lack the ability to replicate it. An example of a fully open-source AI model is Mistral NeMo 12B, while Meta’s Llama 3 serves as an example of a model with just open weights.
Industry experts argue that this initiative could significantly contribute to the democratization of AI technology worldwide. Sean Ren, an associate professor of computer science at the University of Southern California and recognized as Samsung’s AI Researcher of the Year, stated in an interview with CNBC, “This isn’t just a China story. Every time a major lab open-sources a powerful model, it raises the bar for the entire industry.”
Notably, the last time a Chinese open-source AI model was launched (DeepSeek-R1), it negatively affected the stock prices of many major AI firms and had an adverse effect on several US-based companies.
