4 ms·

Everything goes through OpenAI. Different models are used for different purposes, at the moment we use gpt-3.5-turbo and gpt-4-turbo-preview