百度世界 2026：Robin 李彦宏阐述模型与智能体并行演进路线
百度这次把深度思考和微表情识别两款 ERNIE 模型直接开放到千帆平台 API，企业接上就能用。李彦宏讲的思路很有意思：智能体的应用反馈反过来指导模型优化。做企业级应用开发的可以看看千帆上的调用方式。
百度在 Baidu World 2026 大会上宣布了两款新模型开放：ERNIE Light Deep Thinking 模型和 ERNIE Wizshot Micro-Expression 模型，现已通过百度智能云千帆（Qianfan）平台的 API 向企业和合作伙伴提供。李彦宏（Robin）在会上提出的路线是前沿基础模型持续投入与应用层模型、智能体开发同步推进，两者协同演进。
这条路线的关键在于应用需求反向引导模型优化：智能体在实际业务中暴露出的短板，直接决定基础模型的训练和调优方向，而不是模型先发布、应用再适配。微表情识别模型 ERNIE Wizshot 则把能力延伸到多模态细粒度场景。
对企业客户而言，两款模型通过千帆 API 交付，意味着可以直接在百度智能云上调用，无需自建推理基础设施。这与 OpenAI、阿里通义等厂商通过云平台输出模型能力的打法一致，竞争焦点正在从模型发布转向应用与模型的联动迭代速度。
Advancing AI means pushing the boundaries of intelligence and turning those capabilities into value for businesses.
The approach Robin outlined at Baidu World 2026 pairs continued investment in frontier foundation models with the parallel development of models and agents. As they evolve together, application needs guide how the models are optimized.
The ERNIE Light Deep Thinking model and ERNIE Wizshot Micro-Expression model are now available via API on Baidu AI Cloud's Qianfan platform for enterprises and partners.
This report is categorized under AITOP's "Models" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include ERNIE, Baidu, 千帆平台, 多模态.
Source attribution: curated from 百度 AI Baidu (@Baidu_Inc). Cross-source corroboration: 11 independent sources have covered this development (including Google Gemini App, Artificial Analysis, 阿里云 Alibaba Cloud), confirming industry-wide signal strength.
