Chinese tech giant Baidu just dropped a major update to its flagship AI model, and the results are turning heads. ERNIE-5.0-0110 has clawed its way to #8 on LMArena's Text leaderboard, making it the only Chinese model to crack the global top 10. It's now officially ranked ahead of OpenAI's GPT-5.1-High and Google's Gemini-2.5-Pro, a notable shift in the AI capability gap.
The model's math skills are particularly impressive, securing the #2 spot globally for mathematical reasoning, trailing only the unreleased GPT-5.2-High. For a domestic model to outperform nearly every publicly available Western system on complex logical tasks marks a significant moment. Baidu announced on X that ERNIE 5.0 has officially exited its preview phase, a comeback story given that its preview version had plummeted to 24th place on LMArena just last November.
ERNIE 5.0 also demonstrated competitive performance in creative writing, instruction following, and coding, ranking in the top 10 across multiple occupational categories. The technical architecture behind these results follows China’s focus on efficiency over raw power. It uses a roughly 2-trillion-parameter Mixture-of-Experts architecture, activating only a small fraction of experts—just 3% per inference. This design reduces per-query compute compared with dense systems, though it adds engineering complexity. Baidu also trained ERNIE 5.0 natively across text, images, audio, and video, rather than retrofitting multimodal modules onto a language-only core, making it a natively "omni-modal" AI.
The consumer market has been a tougher battleground. Domestic competitors like ByteDance's Doubao command 101 million monthly active users, and DeepSeek's cost-efficient models triggered a price war that forced Baidu to abandon its paid subscription model entirely last April. However, these consumer struggles haven't slowed Baidu's enterprise momentum. ERNIE now powers smart city command centers across China, serves all systemically important Chinese banks, and processes 16.5 billion API calls daily. This B2B stronghold has insulated Baidu from cutthroat consumer competition while funding continued model development.
The UI for the ERNIE chatbot is user-friendly, with pre-customized features for writing, reading, image editing, and general use to simplify prompts. Interestingly, the most recent version does not come with web search enabled, making it purely offline, though users can switch to the previous Ernie 4.5 for updated information with web search. Reactions have been mixed, with some users awaiting a more detailed breakdown of the model’s performance and benchmarks, which the company promised to release soon. It's also worth noting that while ERNIE 5.0 matches GPT-5 and Gemini 2.5 on specific benchmarks, many Western labs have already moved to GPT-5.2 or Gemini 3, making this more of a catch-up than a leap ahead. Users can test the model for free at ernie.baidu.com.
