Remix.run Logo
Remix clone Hacker News
new | show | ask | jobsGithub
▲ petu 3 hours ago

It's larger than previous V4 Flash.

  552B in ~FP4, 306GB.   
  196B of FP8 Engrams, another 204GB, not necessary to keep in RAM.  
  KV cache sees another 4x size reduction, just 900MB for 1M.
So 384GB needed for a chance of achieving useful speeds. Three Sparks or quad RTX PRO 6000.
▲npodbielski 2 hours ago | parent [-]

Or two gorgon halos?