Google Research's Gemini-SQL2 Leads Text-to-SQL Benchmarks
Google Research has developed Gemini-SQL2, an AI model capable of converting natural language into executable SQL queries. Built on Gemini 3.1 Pro, the system has achieved the highest accuracy on the BIRD benchmark, reaching 80.04 percent. This performance reportedly places it ahead of models from OpenAI and Anthropic, with Google suggesting its potential to enhance natural language features across its data services.
Google Research's Gemini-SQL2 has reportedly achieved a leading position in text-to-SQL conversion benchmarks. This system is designed to translate natural language commands into executable SQL queries.
The model, which is built upon Gemini 3.1 Pro, recorded an 80.04 percent accuracy rate on the BIRD benchmark. This score places Gemini-SQL2 significantly ahead of competing models developed by companies such as OpenAI and Anthropic.
Google has indicated that this technology could be leveraged to improve natural language features across its various data services, potentially streamlining interactions with databases.
According to The Decoder AI, Gemini-SQL2's performance marks a notable advancement in the field.
Advertisement
AdSense slot • inline
