Alphane Moon@lemmy.world to Technology@lemmy.worldEnglish · 11 months agoSam Altman said startups with only $10 million were 'totally hopeless' competing with OpenAI, DeepSeek's disruption says otherwisewww.tomshardware.comexternal-linkmessage-square75linkfedilinkarrow-up1507arrow-down118
arrow-up1489arrow-down1external-linkSam Altman said startups with only $10 million were 'totally hopeless' competing with OpenAI, DeepSeek's disruption says otherwisewww.tomshardware.comAlphane Moon@lemmy.world to Technology@lemmy.worldEnglish · 11 months agomessage-square75linkfedilink
minus-squareMojave@lemmy.worldlinkfedilinkEnglisharrow-up33·11 months ago DeepSeek claimed the model training took 2,788 thousand H800 GPU hours, which, at a cost of $2/GPU hour, comes out to a mere $5.576 million. That seems impossibly low. DeepSeek is clear that these costs are only for the final training run, and exclude all other expenses
minus-squareKnock_Knock_Lemmy_In@lemmy.worldlinkfedilinkEnglisharrow-up5·11 months agoThere would have been many other runs before the release version.
minus-squareflamingo_pinyata@sopuli.xyzlinkfedilinkEnglisharrow-up1·11 months agoThis claims only training costs come out to $5M. It doesn’t include previous versions (there had to have been multiple attempts before the final one), or salary of the people working on it.
There would have been many other runs before the release version.
This claims only training costs come out to $5M. It doesn’t include previous versions (there had to have been multiple attempts before the final one), or salary of the people working on it.