Amid the wave of artificial intelligence, Baidu officially launched its latest multimodal large model — ERNIE 5.0 today. With an impressive 2.4 trillion parameters, this model demonstrates extraordinary language understanding and generation capabilities, marking a significant breakthrough in the field of artificial intelligence.

ERNIE 5.0 adopts a brand-new multimodal unified modeling technology, which means it can process various types of information simultaneously, including text, images, audio, and video. This technology enables the model to better integrate and optimize different types of data, thus achieving more accurate understanding and generation capabilities. In over 40 authoritative benchmark tests, ERNIE 5.0 outperformed many international top models, such as Gemini-2.5-Pro and GPT-5-High, firmly securing its position among the global leaders.

In addition, ERNIE 5.0 features an ultra-large-scale mixture-of-experts structure, with an activation parameter ratio of less than 3%, effectively improving inference efficiency while maintaining powerful functions. This innovative design provides users with a smoother experience, especially when dealing with complex tasks, where its response speed and accuracy are significantly enhanced.

To further enhance the professionalism and depth of ERNIE 5.0, Baidu has invited 835 experts from multiple fields, including technology, finance, culture, and education, to guide and calibrate the model. Their contributions in terms of logical rigor, professional depth, and creative quality will help ERNIE 5.0 continue to improve, making it more aligned with user needs and expectations in practical applications.

Currently, individual users can experience this powerful large model through the ERNIE App and the ERNIE Bot website, while enterprises and developers can call it via the Baidu Qianfan platform. Baidu hopes that ERNIE 5.0 will promote the application of artificial intelligence across all industries and support the digital transformation of society.