alt Hacker NewsHow come no other big model seems to be able to deliver the same type of extremely low cache cost though, if their techniques are public?
Replies
jboss10 • yesterday at 4:33 PM
I think the "architectural features" are part of the model, not the kv cache. So implementing it would be difficult and expensive.
petu • yesterday at 4:22 PM
Deepseek V4 paper is just ~three months old
sudosysgen • yesterday at 4:27 PM
Many of these techniques haven't been published very long ago - it often takes a good 6-8 months for techniques to percolate. But also, they come at a complexity cost and, seemingly, also at a stability cost.
➕ show 1 reply
They can and almost certainly are doing similarly impressive engineering works internally.
They just aren't in any hurry to forward those cost savings to you.