EntSQL: leaderboard

Metric: Execution accuracy (%) on the 1,066 English questions with the long-form domain document provided (w/Doc), weighted over the easy, medium and hard SQL-length groups; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 8 models tracked.

Top models

#ModelScore
1Claude Opus 4.613.3
2GLM-5.112.9
3Qwen 3.6 Max Preview12.6
4Claude Sonnet 4.611.6
5GPT-5.411.2
6Gemini 3.1 Pro (Preview)9.9
7Kimi K2.68.3

Interactive version: theaggregate.ai/benchmark?slug=entsql · How It Works · Data refreshed daily, snapshot 2026-09-29.