Google's Gemini-SQL2 achieves 80.04% on BIRD leaderboard

Arxiv CS.CL (NLP/language models)

The text-to-SQL system powered by Gemini 3.1 Pro scored 80.04% execution accuracy on the BIRD Single-Model Leaderboard, surpassing Google's own previous models. The system translates natural language into executable SQL queries, marking a new state-of-the-art.

GoogleGeminibird-benchmark

See 2 more sources

Google Releases Gemini-SQL2: Gemini 3.1 Pro Text-to-SQL Scores 80.04% on BIRD Single-Model Leaderboard3 days agoAsif Razzaq

Google releases Gemini-SQL2, breakthrough text-to-SQL capability model3 days agoBuildwithVigneshDiscuss

Arxiv CS.CL (NLP/language models)·Sai Ashish SomayajulaMarianne Menglin LiuChuan LeiFjona ParllakuDaniel GarciaRongguang WangSyed Fahad Allam ShahAnkan BansalSujeeth BharadwajTao ShengSujith RaviDan Roth

5 days ago