Google Unveils Gemini-SQL2: Achieves 80.04% on BIRD Leaderboard

Google's research team has introduced Gemini-SQL2, a significant advancement in text-to-SQL technology, built on the Gemini 3.1 Pro framework. This new system achieved an impressive 80.04% execution accuracy on the BIRD Text-to-SQL Leaderboard, outperforming the previous version, Gemini-SQL. The BIRD benchmark evaluates whether the generated SQL queries not only appear valid but also execute correctly and return accurate results. Gemini-SQL2 is designed to convert natural language inquiries into executable SQL queries, addressing the complexities involved in accurately generating SQL from everyday language. The announcement highlighted the potential for enhanced SQL understanding to improve various Google data services, although specific product integrations for Gemini-SQL2 have yet to be confirmed. The BIRD benchmark, which includes a vast array of question-SQL pairs from diverse domains, sets a high standard for evaluating text-to-SQL systems. Gemini-SQL2's performance places it in a competitive position, with Google holding the top two spots on the leaderboard. The company continues to refine its technology, aiming to bridge the gap between machine-generated and human-level SQL accuracy.