🧵1/19 DeepSeek lied about it's costs.
— tetsuo.ai - e/acc (@7etsuo) February 4, 2025
DeepSeek R1's training did not actually cost only $6 million, that figure was an incomplete and selective representation of the costs. They also have access to around 50,000 Hopper GPUs.
Here's proof.👇
Tetsuo points out some inconsistencies with Deepseek's training figures, immediately gets dog-piled by CCP shills and bots
https://x.com/7etsuo/status/1886770680794104221
- 21
- 44
Now playing: DK Island Swing - Staff Credits Arrangement (Tropical Freeze).mp3
Jump in the discussion.
No email address required.
Doesn't matter.
Merely the implication that they could do it almost crashed the AI market and set flames under the asses of Altman and other techbro scammers, which shows what a fricking clusterfrick the whole AI hype is.
Jump in the discussion.
No email address required.
Even if it cost more than $6 million, Deepseek's papers still suggest their methods saved them a lot of money. Not like it matters, since these LLMs still are limited in their use cases so far.
Jump in the discussion.
No email address required.
More options
Context
More options
Context