OceanBase leads global data agent benchmark with domestic model

OceanBase's Data Agent solution recently topped the International Data Agent Benchmark (DAB), achieving a 90.62% accuracy rate and ranking first. This marks the first submission to surpass the 90% accuracy threshold on the list. The achievement was made using the domestic large model GLM-5.2, surpassing solutions built on overseas models like GPT, Claude Opus, and Claude Fable. The submission, internally named Scout, will be integrated into OceanBase DataPilot. The DAB, launched by UC Berkeley EPIC Data Lab and Hasura PromptQL, evaluates AI's ability to extract answers from complex data environments. Unlike traditional Text-to-SQL, DAB tests the full capability from data understanding to result verification. OceanBase's system uses DataLens to build data profiles, plan execution paths, and validate results through evidence tracking and answer validation. OceanBase's success highlights the growing role of domestic models and databases in AI data analysis. The company is evolving from a database to an AI data platform, with DataPilot targeting AI-driven data analysis scenarios. This ranking validates OceanBase's shift toward becoming an AI data working engine, not just a data infrastructure.