Baidu To Launch Ernie 4.5 AI With Enhanced Features In March

China’s tech giant Baidu is set to launch the next iteration of its artificial intelligence model, Ernie 4.5, in mid-March, according to sources familiar with the matter. This upgraded model promises significant improvements, particularly in reasoning capabilities, alongside enhanced multimodal features.

The Ernie 4.5 model will be able to process and integrate diverse data formats—such as text, video, images, and audio—enabling it to seamlessly convert content across these mediums. Baidu had previously announced that it plans to gradually roll out the Ernie 4.5 series over the next few months, with an official open-source release slated for June 30.

Baidu declined to comment on the matter when contacted.

The announcement of Baidu’s new AI model comes at a time of heightened competition in the industry, sparked by Chinese startup DeepSeek’s debut of AI models that claim to rival, or even exceed, the capabilities of top-tier U.S. models—at a significantly lower cost. DeepSeek’s success has disrupted the AI landscape, intensifying the global race for AI supremacy.

Despite claiming that its Ernie large language model is comparable to OpenAI’s GPT-4, Baidu has faced challenges in gaining widespread adoption. The emergence of DeepSeek has only intensified these struggles, further complicating Baidu’s aspirations in the AI space.

Baidu CEO Robin Li, who had once argued that keeping AI models closed-source was the key to advancing the field, has shifted his stance in response to DeepSeek’s achievements. During an earnings call earlier this month, Li described Ernie 4.5 as the company’s “best model ever,” acknowledging that DeepSeek’s success had influenced the decision to adopt an open-source approach.

The competitive pressure has been further escalated by Alibaba, which recently announced that it would make its video- and image-generating AI model, Wan 2.1, open source, adding to the growing competition in the sector.