# A New Benchmark Argues Text-to-SQL Accuracy Above 89 Percent Does Not Survive Real Enterprise Schemas

ESQ-Bench builds six populated Oracle schemas with 465 tables and tests for queries that run cleanly and return the wrong answer.

- Published: 2026-08-27T06:20:08.967Z
- Canonical: https://polylog.news/ai/2026-08-27/a-new-benchmark-argues-text-to-sql-accuracy-above-89-percent
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.AI](https://arxiv.org/abs/2608.23569)

Natural language to SQL is one of the few agent capabilities that enterprises have actually deployed at scale, and the reported numbers are strong. The best-performing current systems exceed 89 percent execution accuracy on Spider and BIRD,…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-08-27/a-new-benchmark-argues-text-to-sql-accuracy-above-89-percent (subscription information: https://polylog.news/pricing).