It's placed at 10. Below claude-3.5-sonnet, GPT 4.1 and o3-mini.

yeah this was a surprising result. of course, bear in mind that testing an LLM on SQL generation is pretty nuanced, so take everything with a grain of salt :)

[dead]