Bird-SQL (dev) — Results
| Rank | Model | Score |
|---|---|---|
| 1 | sensenova-loop-9b-preview | 71.38 |
| 2 | gemini-2-0-flash-lite | 57.4 |
| 3 | gemini-2-0-flash | 56.9 |
| 4 | gemma-3-27b | 54.4 |
| 5 | gemma-3-12b | 47.9 |
| 6 | granite-4-2-30b | 41.85 |
| 7 | nemotron-3-super-120b-a12b | 41.8 |
| 8 | granite-4-2-8b | 41.07 |
| 9 | gemma-3-4b | 36.3 |
| 10 | gemma-3-1b | 6.4 |
B
Bird-SQL (dev)
1 phaseActive
Comprehensive text-to-SQL benchmark with 12,751 question-SQL pairs across 95 databases spanning 37+ professional domains.
Overview
Bird-SQL (dev)
Quick answer: BIRD (BIg Bench for LaRge-scale Database Grounded Text-to-SQLs) is a comprehensive text-to-SQL benchmark with 12,751 question-SQL pairs across 95 databases (33.4 GB) spanning 37+ professional domains including healthcare, finance, and sports.
Benchmark Details
| Property | Value |
|---|---|
| Tasks | 12,751 question-SQL pairs |
| Databases | 95 databases, 33.4 GB total |
| Domains | 37+ professional domains |
| Metric | Execution accuracy |
| Challenges | Complex schemas, dirty data, domain knowledge |
Source: Li et al. 2023. Last updated 2026-07-24.