Baidu to Open-Source Ernie AI Model Starting Today, Could Make It China’s 2nd DeepSeek Moment

Baidu, the Chinese tech giant, is making its Ernie artificial intelligence (AI) model open-source on Monday. The company announced the move in February, and has already made its chatbot free to use. This is being viewed as a watershed moment in China's AI space, as this is the first time a major AI player has decided to release its proprietary technology to the open community. As per reports, experts believe Baidu's decision could have a similar impact on the global AI market as DeepSeek-R1 did when it first emerged.

Baidu to Open-Source Ernie in a Gradual Manner

On February 14, Baidu confirmed that its flagship large language model (LLM) Ernie would be made available to the open community. The WeChat post mentioned that the company would open-source its “next generation of AI model,” which many believe to be Ernie 5. Notably, in March, the company released Ernie 4.5 foundation model, alongside its first reasoning model dubbed Ernie X1.

While the company is shifting its strategy to open-source, it is currently not known whether the AI models will be fully open-sourced or partially. The former refers to when a company or a firm releases the model weights, training cookbook, and details about the architecture. This enables developers to not only freely use the models but also modify, replicate, and build upon them.

However, in the latter's case, only the model weights are made open, so the company still holds the other details private. With this, developers can only use and fine-tune the model, but cannot replicate it. An example of a fully open-source AI model is Mistral NeMo 12B, while one with just open weights is Meta's Llama 3.

Experts believe the move could have a positive impact on democratising AI across the globe. Sean Ren, associate professor of computer science at the University of Southern California and Samsung's AI Researcher of the Year, told 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.”

Interestingly, last time a Chinese open-source AI model was released (DeepSeek-R1), it shot down the stock prices of major AI players, and had a net negative impact on most US-based companies.