11 ms·

Runpod, 2xA40. Not sure why you think buying an entire inference server is a necessity to run these models.