Baidu's bmu expands with key talent moves as ernie large model advances

Baidu's Basic Model Research Department (BMU) has recently expanded its team with the addition of experienced researcher Wu Qin from a North American Frontier Lab and Wei Haoran, a former DeepSeek researcher, now leading the ERNIE multimodal algorithm team. These hires come as Sun Tianxiang, BMU head since July 2026, drives upgrades to the ERNIE large model. Baidu has maintained a rapid update pace in foundational models, with DeepSeek integration announced in early 2025. The ERNIE large model 4.5 series was open-sourced by June 2025, followed by an X1.1 release in September, improving factual accuracy by 34.8%. The 5.0 version, launched in November 2025, surpassed 2.4 trillion parameters, marking a full-modal model. In 2026, further talent additions have propelled ERNIE toward new organizational and technological milestones. Baidu's strategic focus on talent and innovation aligns with its efforts to match global AI research trends. The company's continuous updates and expansions highlight its commitment to advancing large model capabilities. With recent hires and model iterations, Baidu aims to strengthen its position in the competitive AI landscape.