Defog (Defog.ai)

Defog is an enterprise text-to-SQL platform built on its open-source SQLCoder LLM family, letting teams turn natural-language questions into SQL and answers. SQLCoder model weights are free on Hugging Face for self-hosting, while a paid cloud/enterprise product adds managed deployment. Defog is a small YC (W23) team (~$2.2M raised) with reported M&A activity in 2025, so confirm current status.

Defog (Defog.ai), a Y Combinator (W23) company, built an enterprise text-to-SQL platform designed to let business and technical users ask questions in natural language and get accurate SQL and answers back. Its most recognizable contribution is SQLCoder, an open-source family of LLMs (released in multiple sizes such as 8B, 14B and 32B) fine-tuned specifically for SQL generation and published on Hugging Face, where the models gained substantial downloads and GitHub stars and were reported to outperform general-purpose models on Defog's internal SQL benchmarks. The commercial product wraps this capability with cloud and self-hosted deployment options for teams that need accurate, governed natural-language querying over their databases, including those with strict data-residency requirements who prefer to run the open models themselves. This dual approach - open-source models plus a paid enterprise product - is Defog's defining characteristic. Defog is a small team (headquartered in Singapore) that raised a reported ~$2.2M seed round via Y Combinator and other investors. It has historically reported modest revenue with a lean team, and there have been reports of M&A activity around the company in 2025, so its independent status may be in flux. Enterprise pricing has been reported around the low thousands per month for unlimited cloud usage, while the SQLCoder weights remain free to self-host. Buyers should confirm current company status and commercial terms directly.