reasoning benchmark
Bird-SQL (dev)
BIRD (BIg Bench for LaRge-scale Database Grounded Text-to-SQLs) is a comprehensive text-to-SQL benchmark containing 12,751 question-SQL pairs across 95 databases (33.4 GB total) spanning 37+ professional domains. It evaluates large language models' ability to convert natural language to executable SQL queries in real-world scenarios with complex database schemas and dirty data.
Updated Aug 11, 2026
Models7
Model coverage7
MetricScore
EvidenceB
Bird-SQL (dev) Ranking
Higher score ranks better on this benchmark.
| 1 | GO | 57.4% | 100.0% | 7 | C | |
| 2 | GO | 56.9% | 83.3% | 7 | C | |
| 3 | GO | 54.4% | 66.7% | 7 | C | |
| 4 | GO | 47.9% | 50.0% | 7 | C | |
| 5 | NV | 41.8% | 33.3% | 7 | C | |
| 6 | GO | 36.3% | 16.7% | 7 | C | |
| 7 | GO | 6.4% | 0.0% | 7 | C |
Bird-SQL (dev) Score Distribution
A closer view of the leading scores on this benchmark.
Bird-SQL (dev) Highlights
The leading models and scores on this benchmark.
What is Bird-SQL (dev)?
What Bird-SQL (dev) measures and how its scores work.
BIRD (BIg Bench for LaRge-scale Database Grounded Text-to-SQLs) is a comprehensive text-to-SQL benchmark containing 12,751 question-SQL pairs across 95 databases (33.4 GB total) spanning 37+ professional domains. It evaluates large language models' ability to convert natural language to executable SQL queries in real-world scenarios with complex database schemas and dirty data.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- Bird-SQL (dev)
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- bird-sql-(dev)|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about Bird-SQL (dev).
Which model scores highest on Bird-SQL (dev)?
Gemini 2.0 Flash-Lite is currently ranked first with 57.4%.
What does Bird-SQL (dev) measure?
BIRD (BIg Bench for LaRge-scale Database Grounded Text-to-SQLs) is a comprehensive text-to-SQL benchmark containing 12,751 question-SQL pairs across 95 databases (33.4 GB total) spanning 37+ professional domains. It evaluates large language models' ability to convert natural language to executable SQL queries in real-world scenarios with complex database schemas and dirty data.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
7 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.