Wenxin Large Model User Scale Reaches 45 Million, Over 500 Plugins


After Sun Tianxiang assumed his position in July, Baidu Basic Model R&D Department has initiated a new round of organizational and technical restructuring around the Wenxin Large Model. Wei Haoran, former researcher at DeepSeek, joined with his team and was appointed as the head of Wenxin Multimodal Algorithm. Baidu continues to recruit core technology experts to strengthen large model development.
Baidu's Basic Model R&D Department (BMU) has completed key talent deployment: On September 3rd, the head Sun Tianxiang announced that senior researcher Wu Qin was recruited from a North American Frontier Lab to enhance the pre-training capabilities of the Wenxin Large Model; former DeepSeek researcher Wei Haoran has officially transferred to BMU as the head of Wenxin Multimodal Algorithm. His team previously open-sourced the Unlimited OCR model.
On June 25, Baidu Wenxin Yanyan released an upgrade notice, announcing the deep integration of the platform architecture and service entry points. Starting from midnight on June 25, 2026, the official website question-answering entrance will be upgraded. Users will need to use AI functions such as intelligent dialogue, creative writing, office assistance, and information queries, and experience the latest Wenxin large model through the new entrance.
Baidu recently established the Baidu Model Committee (BMC) as the top decision-making and coordination body for its large model strategy, composed of young researchers. The basic and applied model R&D departments report directly to it, enabling unified management from tech development to product deployment, marking a shift to systematic competition in large models.....
Baidu released the new generation Wenxin large model ERNIE-5.0-0110, ranking eighth with 1460 points on the LMArena global text ranking list, making it the only Chinese domestic large model to enter the top ten. Its ability to handle mathematical problems is particularly outstanding, rising to the second globally,仅次于GPT-5.2-High.