Baidu Intelligent Cloud has launched a workspace called Qianfan AppBuilder, which enables anyone to develop AI native applications. It consists of two layers of services: components and frameworks. "Components" are AI capability components that encapsulate underlying service capabilities, such as text recognition and image generation, large language model capabilities like long text summarization and nl2sql, and basic components like vector databases and object storage. Each component has a specific function. "Frameworks" are AI native application frameworks that connect and combine different components. They can accomplish specific tasks in certain scenarios. For example, AppBuilder provides frameworks such as Retrieval-Augmented Generation (RAG), Intelligent Agent, and Intelligent Data Analysis (GBI). Qianfan AppBuilder has two product forms: code-based and low-code-based. The code-based form is suitable for users with deep AI native application development needs, providing SDKs, development environments, debugging tools, sample code, and application components. The low-code-based form is suitable for users without programming foundations, offering visual tools that allow users to quickly customize and launch AI native applications through point-and-click operations. To address challenges in large model development, such as difficulties in fine-tuning models, incomplete model evaluation, and insufficient computing and inference resources, Baidu Intelligent Cloud has upgraded the Qianfan Large Model Platform, focusing on "efficiency improvement" and "cost reduction." After the upgrade, the Qianfan platform has pre-installed the largest number of 54 basic and industry-specific large models in China, enhancing their model capabilities. Additionally, the platform has introduced functions such as data statistics analysis and data quality inspection, along with a data cleaning visualization pipeline, enabling the construction of high-quality large model data fuel. Qianfan platform is also the first MaaS service platform in China that supports multi-dimensional data analysis for large models. Furthermore, the platform has introduced an automated + manual dual-model evaluation mechanism, significantly improving the efficiency and quality of model evaluation. The Qianfan platform also provides flexible pricing methods, such as batch computing, to meet various business scenario demands of customers and help enterprises utilize large models at a low cost. The Qianfan platform has served over 40,000 enterprise users, assisting them in fine-tuning nearly 10,000 large models. By training large models using the Qianfan platform, costs can be reduced by up to 90%.