Google Research Posts Gemini-SQL2 at 80.04% on the BIRD Text-to-SQL Single-Model Leaderboard

LLM-Stats·low signal

Google Research's Gemini-SQL2, powered by Gemini 3.1 Pro, reportedly posted 80.04% execution accuracy on the BIRD single-model text-to-SQL leaderboard. If it holds up, it's a notable jump for production analytics and natural-language-to-query tooling. Single-source so far, so treat the exact figure as preliminary until a second confirmation lands.