Available on Hugging Face and released under the Apache 2.0 license, Gretel’s dataset consists of over 100,000 high-quality synthetic Text-to-SQL samples with SQL metadata and spans 100 verticals.