How Baidu’s Wenxin Model and PaddlePaddle Accelerate AI in the Intelligent Era

On April 24, the 3rd China International Software Development Conference and the 40th anniversary of the China Software Industry Association were held in Beijing, where 32 distinguished figures in the Chinese software industry received honor medals.

Keynote by Baidu CTO Wang Haifeng

Wang delivered a speech titled “Operating Systems in the Intelligent Era,” emphasizing how deep‑learning frameworks combined with large models are reshaping operating system functions.

PaddlePaddle and Wenxin Model Highlights

The PaddlePaddle deep‑learning platform, China’s first domestically developed industrial‑grade open‑source platform, includes a core framework, model library, development toolkit, and a community for learning and training. The Wenxin large model, an innovative knowledge‑enhanced model, integrates trillions of data points and hundreds of billions of knowledge items, achieving capabilities such as knowledge‑enhanced retrieval, dialogue, and code generation.

Joint optimization of PaddlePaddle and Wenxin has raised the weekly training efficiency of the Wenxin model to 98.8%, a 5.1‑fold improvement over the previous year’s Wenxin‑Yiyan release, and boosted inference speed by 105 times. The ecosystem now comprises 12.95 million developers, serves 244 000 enterprises, and hosts 895 000 models.

Intelligent Agents and Code Assistants

Wang identified intelligent agents as a major future direction, describing them as models built on foundational models that undergo supervised fine‑tuning, preference learning for decision‑making, and reinforcement learning for result reflection, enabling them to read manuals, learn tool usage, and execute tasks.

Leveraging Wenxin’s natural‑language and code abilities, Baidu has created a code‑intelligent agent and an intelligent code assistant. The code agent aims to let anyone perform tasks traditionally reserved for programmers, while the assistant helps professional developers write higher‑quality code more efficiently.

AI as a Driver for Technological Self‑Reliance

Wang concluded that artificial intelligence is a pivotal force in the new wave of technological revolution and industrial transformation, urging accelerated AI development to contribute to high‑level technological self‑reliance.