With the rapid advancement of AI large language models and computing power infrastructure, Ant Digital Technologies’ open-source initiative directly addresses industry pain points, empowering ordinary users to efficiently handle business data using everyday language. The first released real-time Text-to-SQL framework significantly improves text-to-database interaction efficiency, with its core strength lying in AI-driven semantic understanding—laying the foundation for developers to quickly build query solutions. During trial operations, this technology achieved an accuracy rate exceeding 92% at a city commercial bank, a threefold improvement over traditional approaches, highlighting AI’s vast potential in optimizing computing power resources.
The global business intelligence market is experiencing explosive growth, with projected market size reaching $47.48 billion by 2025. China’s market, with a compound annual growth rate of 12.7%, is expected to hit $1.79 billion by 2028. This expansion underscores the urgent demand for efficient data processing infrastructure within the AI industry, and Ant Digital Technologies’ technology is key to addressing this challenge. In the future, the solution will integrate modules for database comprehension, industry knowledge mining, and real-time multi-turn interactions, comprehensively enhancing AI’s application capabilities in complex scenarios and offering new pathways for constructing industry knowledge graphs.
In the authoritative BIRD-SQL benchmark, Agentar SQL surpassed global giants like Google, becoming the industry leader. The benchmark covers 37 real-world scenarios across finance, electricity, and healthcare, featuring complex tasks and massive data volumes, demonstrating the advantages of AI large language models in parsing intricate data structures. However, Zhang Peng, technical lead at Ant Digital Technologies, noted that real-world applications face multiple challenges such as ambiguity in spoken language and integration of specialized knowledge—far beyond what simple models can handle. He emphasized that building a complete capability system is essential for industrial usability, involving deep integration between large models and databases, as well as AI’s self-evolution capability, driving data processing toward intelligent transformation.
The open-source release of Agentar-Scale-SQL has been published on platforms like arXiv and GitHub, attracting global developer attention. This move not only accelerates the adoption of AI technology in enterprise applications but also lowers innovation barriers through shared infrastructure, highlighting the critical role of chips and computing power resources in AI industry collaboration. Looking ahead, Ant Digital Technologies plans to continue open-sourcing more comprehensive modules, further advancing the democratization of intelligent data analytics and facilitating the industry’s transition toward an AI-driven future.