3 ms·

I assume there's no reason to use GPT-4-turbo for API calls, as this one is supposedly better and 2x cheaper.