4 ms·

I’ve found input tokens is around 5x more than output, so a task could be a couple million thinking tokens and then a few couple 100k output tokens?