| ▲ | nullsanity 19 hours ago | |
How? A full server of GH200s is half a mil. That runs GLM5.3 Prime and Delegate Flash models. Half a million, in 2026, is a lot for me and you, but pennies for even a modest sized business. And this is before prices go down. Ive been using 9b models on my 4090 for coding and command tasks for over a year now. It's great! You obviously can't be like "Use huge framework and context, and just make me an app" but you can be like "Employee said vpn is running slow. Follow the process documented here: X to diagnose and propose a fix" where X is docs written by a Chinese model for 10 cents. | ||