Group 1 - The 2025 Baidu Cloud Intelligence Conference showcased the launch of the Baidu Intelligent Cloud Qianfan Platform 4.0, marking a new era of enterprise-level services centered around Agents [1] - The Qianfan Platform 4.0 has significantly expanded its model library to include over 150 models, featuring Baidu's self-developed "Steam Engine" video generation model and industry-specific models like Qianfan Huijin for finance [1] - The platform introduced a new RFT (Reinforcement Feedback Tuning) toolchain, providing developers with unprecedented model fine-tuning precision [1] Group 2 - The Qianfan Agent Service Platform achieved breakthroughs with its upgraded multi-modal RAG (Retrieval-Augmented Generation) functionality, enhancing information processing and generation capabilities [3] - The Qianfan Enterprise-level MCP service has been fully upgraded, integrating Baidu's unique AI search MCP Server and other core components to offer robust and flexible service solutions for enterprises [3] - The Qianfan Data Intelligence Service Platform underwent a comprehensive overhaul, providing one-stop multi-modal data management and processing capabilities to support efficient handling and analysis of large volumes of data for precise decision-making [3] Group 3 - Qianfan 4.0 implemented full-link detection of key indicators, links, and logs, offering enterprises comprehensive operational monitoring tools to identify and resolve issues promptly, ensuring stable business operations [3] - This upgrade highlights Baidu's leading position in the intelligent cloud service sector, delivering unprecedented convenience and efficiency to developers and enterprise users [3]

Core Insights - Baidu announced the launch of its latest cloud service, Qianfan 4.0, at the 2025 Baidu Cloud Intelligence Conference, featuring over 150 model services that allow users to access state-of-the-art (SOTA) models immediately [1] Group 1 - Baidu's new video generation model, Steam Engine 2.0, along with more than ten vertical industry models, has been integrated into the Qianfan large model [1] - The company introduced the RFT toolchain, which enhances training efficiency by 43% through the provision of a reward model [1]