.svg){kind=spacer}
Details
- On November 13, 2025, Baidu introduced ERNIE 5.0, a foundational model engineered for seamless integration across text, image, and audio inputs and outputs.
- The model demonstrates capabilities in visual understanding, image generation, and audio comprehension, as shown by live demos.
- Baidu asserts improved creative writing, instruction following, and factual reasoning over its previous ERNIE 4.0 release from 2023.
- Two trial versions—ERNIE 5.0 Preview and Preview 1022, the latter focused on text accuracy—are now available for public experimentation.
- A public test platform is open, with enterprise access and API options in development, but detailed pricing and technical specs have not yet been revealed.
- Baidu affirms ongoing investment in building larger, more powerful models to drive advances in artificial intelligence capabilities.
Impact
Baidu’s ERNIE 5.0 signals mounting competition with OpenAI’s GPT-4o and Google’s Gemini in the push for unified multimodal AI. Its native handling of text, vision, and audio could streamline development and compliance for Chinese enterprises under new AI regulations. The release also puts pressure on domestic rivals like Alibaba and Tencent to quickly match Baidu’s advancements or risk losing ground in the enterprise market.
