Baidu's ERNIE 5 AI Model Rises Up the Rankings—A Math Wiz That Beats OpenAI's GPT 5.1
Baidu’s ERNIE 5.0 AI model scored 1,460 on LMArena’s Text leaderboard, ranking #8 globally and as the only Chinese model in the top 10, ahead of OpenAI’s GPT-5.1-High and Google’s Gemini-2.5-Pro. It is also #2 worldwide in mathematical reasoning, second only to GPT-5.2-High. ERNIE 5.0 exited preview and now delivers notable performance in creative writing, coding, and instruction following, ranking in the global top 10 across several professional fields. The model utilizes a 2-trillion-parameter Mixture-of-Experts architecture, activating just 3% per inference for efficiency and enabling native multimodal capabilities across text, images, audio, and video. Despite losing consumer market share to rivals like ByteDance, Baidu has maintained strong enterprise adoption; ERNIE powers major Chinese banks and smart cities, processing 16.5 billion daily API calls. The model’s user-friendly interface offers specialized sections for different tasks but lacks live web search, which is available in ERNIE 4.5. While ERNIE 5.0 closes the capability gap with leading Western models, Western labs have already advanced to newer versions. Public testing is available at ernie.baidu.com.
