Hacker News new | ask | show | jobs
by bigyabai 202 days ago
>90% of inference hardware is faster if you run an MOE model.