Baidu, Inc. (百度) is a leading Chinese technology giant founded in 2000 by Robin Li and Eric Xu. The company is incorporated in the Cayman Islands with its operational headquarters in Beijing's Haidian District. Over two decades, the company has evolved from a classic search service into a leading developer of platforms and solutions in artificial intelligence, playing a significant role in the development of the national AI ecosystem.
Baidu started as a Chinese analogue to Google, but in 2014, following the appointment of Dr. Andrew Ng as Chief Scientist, it set a course to transform into an AI company. The first major step was the creation of PaddlePaddle (2016) — China's first open-source deep learning framework, which became the technological foundation for all subsequent developments.
In 2019, Baidu introduced ERNIE 1.0 (Enhanced Representation through kNowledge IntEgration) an architecture that integrates knowledge graphs directly into the pre-training process of language models. ERNIE 2.0 (2019) added continual multi-task learning, and ERNIE 3.0 (2021) unified text understanding and generation, achieving state-of-the-art results in 54 Chinese NLP tasks. The ERNIE 3.0 Titan model with 260 billion parameters cemented Baidu's position as a serious player in the field of large language models.
Baidu's primary scientific contribution to the development of LLMs has been the knowledge enhancement approach: the integration of structured knowledge during the pre-training phase by masking entities and phrases, rather than individual tokens. This allowed ERNIE models to better understand semantic relationships and context, especially in logographic languages like Chinese.
Historically, Baidu adhered to a proprietary approach, but the emergence of DeepSeek radically changed its strategy. In February 2025, CEO Robin Li announced plans to open-source the ERNIE models, and on June 30, 2025, the company released the ERNIE 4.5 family under the Apache 2.0 license.
Thus, Baidu occupies a unique position in the industry – it conducts fundamental research combined with the release of successful commercial products and open-source solutions. The company is one of the few Chinese tech giants offering a full AI stack: from chips and cloud infrastructure to foundation models and applications.