2 ms·>90% of inference hardware is faster if you run an MOE model.by bigyabai 1y ago>90% of inference hardware is faster if you run an MOE model.