ERNIE X1.1

Baidu’s ERNIE X1.1 is a next-generation reasoning model focused on higher correctness and controllability. Announced at the 2025 Wave Summit (Sept 9), Baidu reports significant gains over X1: +34.8% factuality, +12.5% instruction following, and +9.6% agentic capability. Media covering the launch also highlighted Baidu’s parity claims with top frontier models.
Under the hood, ERNIE X1.1 builds on the ERNIE 4.5 foundation and introduces an iterative hybrid reinforcement-learning training pipeline to strengthen multi-step reasoning and tool-use behaviors.
Access is broad: you can try it in ERNIE Bot (Wenxiaoyan) and call it programmatically through Baidu Smart Cloud’s Qianfan platform/SDK. Third-party guides and docs provide quick starts for API integration.
Positioning: Baidu pitches X1.1 for complex analysis, coding help, and agent workflows; coverage notes claims of performance on par with leading models, though such comparisons come from launch materials and media reports.