Chinese tech giant Baidu has announced that it is opening up the source code of its generative language model Ernie. This will be the biggest discovery from the Chinese AI industry since the release of DeepSeek and could significantly change the situation in the global market. This is reported by CNBC.
Although Baidu had previously opposed opening up its models, the success of DeepSeek has forced the company to reconsider its strategy. According to Omdia analyst Liang Jie Su, Baidu's move to open source signals a recognition of the effectiveness of open models in competing with proprietary solutions.
"Every time a major lab open-sources a powerful model, it raises the bar for the entire industry," said University of Southern California professor Sean Ren.
Ernie X1 is estimated to be capable of delivering performance on par with DeepSeek R1, but at half the price. Analysts believe this could spark a new "price war," particularly against paid models from OpenAI and Anthropic.
Baidu emphasized that opening up the model's code will help developers around the world build products faster without cost or tool constraints. However, experts remind that open source transparency does not guarantee trust, as it is not known what data the model was trained on and whether the company received consent from the authors.
Opening up Ernie's code could affect the positions of not only Chinese but also Western developers. OpenAI CEO Sam Altman said the company will open source its own model by the end of the summer.