Google Releases Gemini-SQL2: Gemini 3.1 Pro Text-to-SQL Scores 80.04% on BIRD Single-Model Leaderboard

Google Research has announced a significant breakthrough in artificial intelligence-driven database management through the release of Gemini-SQL2, a text-to-SQL system powered by Gemini 3.1 Pro. The new capability demonstrates substantial progress in natural language processing by achieving an 80.04% execution accuracy rating on the BIRD single-model leaderboard, positioning it among the industry's most capable systems for converting human language into structured database queries.

Announced on June 12, 2026, Gemini-SQL2 represents a notable advancement in the text-to-SQL domain, where AI systems interpret natural language instructions and generate corresponding SQL commands for database queries. The BIRD leaderboard measures execution accuracy—a metric that evaluates whether generated queries produce correct results when executed against actual databases, rather than merely checking syntactic correctness. This 80.04% accuracy threshold demonstrates Gemini-SQL2's ability to handle complex database interactions across diverse schemas and query requirements, making it a competitive contender in enterprise database automation scenarios.

The development of Gemini-SQL2 carries several important implications for technology sectors reliant on database management:

• Democratization of data access through reduced technical barriers for non-technical users querying databases
• Enhanced business intelligence capabilities by enabling faster, more accessible data exploration
• Potential cost reduction in database administration and query optimization workflows
• Improved handling of complex, multi-table queries that traditionally required specialized SQL expertise
• Competitive pressure on enterprise software providers to integrate similar AI-driven query generation capabilities

Gemini-SQL2's strong leaderboard performance signals Google's commitment to advancing practical AI applications that address real business challenges. The ability to reliably convert natural language into executable database queries at 80% accuracy rates represents a tipping point where organizations may increasingly adopt such systems for routine data access tasks. As AI continues evolving in database automation, stakeholders across analytics, business intelligence, and enterprise software must prepare for broader adoption of natural language interfaces to data infrastructure. This development underscores how specialized AI capabilities are moving beyond academic benchmarks toward practical deployment in production environments, reshaping how organizations interact with their most critical data assets.