Apparently, on cerebras.ai it runs even faster, like 3k tokens/second

Can’t recall all of the top of my head, but I think they told me they love reverse proxy and a few other features.

Ah, yes. My team and I actually use Nginx extensively with our GeoDjango projects. Find it somehow more flexible.

Thank you for sharing, Chris.

Hey, just out of curiosity, why Python 3.10 and not a bit newer, like .12? Looking to learn from your perspective.