4 ms·

I dont get it, so for slow inference its: hermes > pi > opencode ?? or is hermes better in general? habent tried it hermes/pi yet