.svg){kind=spacer}
Details
- Baidu introduced ERNIE X1.1, the next-generation AI model, at Wave Summit 2025 in Beijing, building on the capabilities of ERNIE X1.
- The company claims a 34.8 percent improvement in factual accuracy, addressing the challenge of hallucinations in large language models.
- Enhanced instruction-following logic enables the model to reject misleading prompts and deliver more precise responses to valid queries.
- ERNIE X1.1 now features computer-vision abilities, accurately analyzing uploaded images and generating detailed caption analyses.
- Live demos highlighted the model's ability to tackle complex, multi-step tasks and simulate particle physics within a rotating 3D environment, showcasing advanced scientific reasoning.
- The model targets enterprise APIs, Baidu's Qianfan AI platform, and is planned for upcoming integration into Baidu Search and Apollo in-car systems.
- No pricing has been released, but early-access documentation is open to Wave Summit attendees and Chinese cloud users starting today.
Impact
Baidu’s ERNIE X1.1 raises the stakes for competitors like Alibaba’s Qwen2 and OpenAI’s expected GPT-5 by sharpening focus on factual reliability—a top concern for enterprise clients. With enhanced multimodal and reasoning skills, the model is well positioned to drive AI adoption in regulated industries and shift cloud spending toward Baidu’s ecosystem in China.
