☆ Yσɠƚԋσʂ ☆ to US NewsEnglish · 3 months agoSo OpenAI strategy was to become too big to fail and live off government bailouts. Nothing says free market like a public safety net for private billion-dollar bets. A beautiful capitalist fairy tale.www.barrons.comexternal-linkmessage-square13linkfedilinkarrow-up150arrow-down10cross-posted to: news@hexbear.netusa@lemmy.ml
arrow-up150arrow-down1external-linkSo OpenAI strategy was to become too big to fail and live off government bailouts. Nothing says free market like a public safety net for private billion-dollar bets. A beautiful capitalist fairy tale.www.barrons.com☆ Yσɠƚԋσʂ ☆ to US NewsEnglish · 3 months agomessage-square13linkfedilinkcross-posted to: news@hexbear.netusa@lemmy.ml
minus-squarePeeOnYou [he/him]linkfedilinkarrow-up2·3 months agothere is no way you’re running the full deepseek model on a $2000 gpu much less computer
minus-squareCriticalResist8Alinkfedilinkarrow-up6·3 months agohttps://digitalspaceport.com/how-to-run-deepseek-r1-671b-fully-locally-on-2000-epyc-rig/ it’s even better, this rig doesn’t even have a GPU.
minus-squarePeeOnYou [he/him]linkfedilinkarrow-up1·2 months agothat is mind blowing! without a gpu at all and getting even 3 t/s? just… wow
minus-squareCriticalResist8Alinkfedilinkarrow-up2·2 months agoAnd it’s only going to get faster… (Note that this is the Q4 model, but otherwise still the full 671b params and thinking mode enabled)
minus-squarePeeOnYou [he/him]linkfedilinkarrow-up2·2 months agooh i missed that… yeah Q4 makes way more sense
there is no way you’re running the full deepseek model on a $2000 gpu much less computer
https://digitalspaceport.com/how-to-run-deepseek-r1-671b-fully-locally-on-2000-epyc-rig/ it’s even better, this rig doesn’t even have a GPU.
that is mind blowing! without a gpu at all and getting even 3 t/s? just… wow
And it’s only going to get faster…
(Note that this is the Q4 model, but otherwise still the full 671b params and thinking mode enabled)
oh i missed that… yeah Q4 makes way more sense