3 ms·

When I was going through this it was because OpenAI had defaulted to /fast mode with 2x token usage