BIRD-History: leaderboard
BIRD-History (Zhejiang University, 2026) contains 1,393 text-to-SQL tasks over 11 databases whose questions rely on business logic and data conventions that are not in the schema or the question but appear in historical SQL query logs. This board is execution accuracy in percent for foundation models generating SQL without a history retriever; higher is better.
Metric: Execution accuracy (%; 1,393 underspecified text-to-SQL tasks over 11 databases whose implicit business knowledge appears only in historical SQL logs; foundation model generating SQL directly without a history retriever, temperature 0). Source: arxiv.org. Saturation forecast: Around December 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 33.24 | #164 |
| 2 | GPT-5.3 Codex | 28.93 | #116 |
| 3 | DeepSeek V3.2 | 21.11 | #406 |
| 4 | Llama 3.3 70B Instruct | 16.65 | #1357 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=bird-history · How It Works · Data refreshed daily, snapshot 2026-11-13.