Baidu Launches ERNIE X1.1, Surpasses DeepSeek R1 and Matches GPT-5, Gemini 2.5 Pro
Users can already access the model through ERNIE Bot, Baidu’s Wenxiaoyan app, and via API on its Model-as-a-Service (MaaS) platform Qianfan.
China's Baidu has introduced ERNIE X1.1, its latest reasoning model, at the Wave Summit 2025, promising significant improvements in factual accuracy, instruction following, and agentic performance.
This follows Alibaba's unveiling of its most powerful large language model yet, Qwen3-Max-Preview (Instruct), boasting more than 1 trillion parameters.
According to Baidu, ERNIE X1.1 reduces hallucinations and delivers stronger reasoning and agentic capabilities compared to its predecessor, ERNIE X1. Benchmark tests showed notable gains: factual accuracy improved by 34.8%, instruction following by 12.5%, and agentic capabilities by 9.6%.
ERNIE X1.1 just launched!
— Baidu Inc. (@Baidu_Inc) September 9, 2025
Unveiled at Wave Summit 2025, our latest reasoning model reduces hallucinations, improves instruction following, and delivers strong agentic capabilities.
✅ Achieves significant performance gains over ERNIE X1, with factual accuracy up 34.8%,… pic.twitter.com/6LkjZqvRS0
The model has also proven competitive against leading AI systems. Baidu said ERNIE X1.1 outperformed DeepSeek R1-0528 and delivered performance on par with OpenAI’s GPT-5 and Google’s Gemini 2.5 Pro.
Built on the foundation of ERNIE 4.5, X1.1 incorporates extensive mid-training and post-training refinements, including end-to-end reinforcement learning, to boost reliability and reasoning depth.
Users can already access the model through ERNIE Bot, Baidu’s Wenxiaoyan app, and via API on its Model-as-a-Service (MaaS) platform Qianfan.
The launch comes as China’s top AI companies accelerate innovation in generative AI and reasoning models to compete globally, while focusing on reducing hallucinations and improving real-world usability for enterprise and consumer applications.
Comments ()