Citing data from third-party AI benchmarks, Baidu chief technology officer Wang Haifeng said at a developer conference in Beijing that the firm’s X1.1 reasoning model had surpassed the performance of DeepSeek-R1, while it matched OpenAI’s GPT-5 and Google’s Gemini 2.5 Pro.
“From reasoning improvements to slow, deep thinking and multimodal abilities, [our] model’s performance and efficiency have increased,” Wang said.
The X1.1 model showed a 34.8 per cent improvement in knowledge accuracy, as well as enhanced agentic capabilities and instruction compliance, according to Wang. He attributed that development to a “mixed reinforcement learning” process.
The closed-source X1.1 model is now available to corporate clients via Baidu’s cloud computing platform. Individual users can access the model through Baidu’s Ernie Bot website and app.
Baidu’s product update reflects heightened competition in China’s AI market, where start-up DeepSeek and Alibaba Cloud have made major strides in open-source models. Alibaba Cloud is the AI and cloud computing services arm of Alibaba Group Holding, owner of the South China Morning Post.
