Wenxin Large Model X1.1 Deep Thinking Model Released

On September 9th, at the WAVE SUMMIT Deep Learning Developers Conference 2025, Baidu officially released the Wenxin Large Model X1.1 deep thinking model. The Wenxin Large Model X1 is a deep thinking model trained based on the Wenxin Large Model 4.5, and the upgraded X1.1 mainly adopts an iterative mixed reinforcement learning training framework. On one hand, through mixed reinforcement learning, it improves the effects of both general tasks and intelligent agent tasks; on the other hand, through iterative production and training of self-distillation data, the overall effectiveness of the model is continuously improved. Compared to Wenxin X1, X1.1 has increased factual accuracy by 34.8%, instruction compliance by 12.5%, and intelligence by 9.6%. Currently, the Wenxin Large Model X1.1 has been officially launched on Baidu Intelligent Cloud Qianfan platform, and is fully open for use by enterprise customers and developers.