News

Baidu Open-Sources Ernie AI Model, Pressuring Global Pricing Structures

• By John K. Waters
• 06/29/2025

China's leading search engine provider Baidu said on Sunday it will open-source its flagship generative AI chatbot, Ernie, marking a major strategic shift for the firm and intensifying global pressure on AI model pricing.

The move, which begins with a gradual rollout starting today, makes Baidu the latest Chinese tech giant to embrace open-source strategies in an effort to challenge Western dominance in AI. The company had previously adhered to a closed-source, proprietary model but now joins a growing group of firms using open-source to gain market share and foster broader developer adoption.

The Ernie model, developed by Baidu's AI unit, will be released with full access to its underlying code, allowing developers worldwide to customize and deploy it freely.

The open release is expected to expand Baidu's developer ecosystem and serve as a counterweight to the paid-access APIs offered by U.S.-based competitors like OpenAI and Anthropic.

"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," said Sean Ren, an associate professor of computer science at the University of Southern California, in a statement.

Analysts interpret the announcement as an aggressive bid by Baidu to accelerate adoption of its models while reshaping AI market dynamics. The firm's monthly active users for Ernie remain modest at 23 million, far behind ByteDance's rival chatbot Duobao, which has 83 million. Baidu's API share also trails behind DeepSeek's, according to Omdia.

In earlier signs of this shift, Baidu had already made Ernie free to use and followed with the release of its Ernie 4.5 and X1 models at reduced costs. The "Turbo" variants of both were introduced shortly after, slashing prices by as much as 80%.

The announcement follows a series of pricing disruptions by competitors such as DeepSeek Ltd., which has led to downward pressure on AI service pricing across both U.S. and Chinese markets. Baidu may also be attempting to capitalize on DeepSeek's reported delays in training its next-generation model, R2, amid U.S. export sanctions affecting access to Nvidia GPUs.

Open-source AI development has drawn scrutiny from U.S. lawmakers concerned about national security and potential misuse. Earlier this year, the U.S. House Select Committee on the Chinese Communist Party labeled Chinese open-source AI efforts a security risk.

Still, Baidu's strategy reflects a broader trend: the commoditization of high-performance AI models. With open weights, customizable architectures, and no access fees, open-source offerings are gaining traction—particularly in global markets where infrastructure flexibility and local language support are priorities.

Baidu's move is likely to intensify the debate over pricing and control in the AI sector. It may also prompt further open-source announcements from Western AI labs seeking to preserve developer mindshare.

OpenAI CEO Sam Altman acknowledged the challenge in January, stating the company must adapt its strategy as its performance advantage narrows. Altman has since confirmed plans to release an open-weights model.

As AI technology matures, Baidu's pivot reflects an industry-wide recalibration. Proprietary performance may no longer be the differentiator it once was—pricing, openness, and developer reach are now shaping the new competitive frontier.