LiveSQLBench: leaderboard
Dynamic contamination-free text-to-SQL benchmark for real-world database tasks, including business-intelligence queries, CRUD/management SQL, hierarchical knowledge bases, and large industrial-scale database variants.
Metric: Success Rate (self-reported). Source: benchmarklist.com. Status: saturation imminent. 33 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 43.1 |
| 2 | Claude Opus 4.6 | 39.43 |
| 3 | GPT-5.5 (xHigh) | 37.36 |
| 4 | Kimi K2.6 | 36.43 |
| 5 | GLM-5.1 | 35.29 |
| 6 | Qwen 3.6 Max | 33.79 |
| 7 | GPT-5.4 | 33.56 |
| 8 | GPT-5.3 Codex | 33.33 |
| 9 | GPT-5 | 31.15 |
| 10 | O3 Mini | 31.15 |
| 11 | Claude Sonnet 4.5 | 30.46 |
| 12 | Kimi K2.5 | 29.89 |
| 13 | O3 | 29.54 |
| 14 | O4 Mini | 29.54 |
| 15 | Claude Sonnet 4 | 27.01 |
Interactive version: theaggregate.ai/benchmark?slug=livesqlbench · How It Works · Data refreshed daily, snapshot 2026-09-05.