@drq yep. "A bit slow, but okay" is the best option we have now, I guess. And future seems even darker.
Running something like 80B models published lately or some behemoths like Grok with its 314B parameters seems nearly impossible for most enthusiasts 🤷🏻♂️
@rayslava Actually, 8b runs fine, like ChatGPT level fine when it comes to speed. Same difference.
70b is much slower, but aside from that still does run okay. I'll try to experiment with more powerful and modern hardware, and let you know.