←back to thread

Deploying DeepSeek on 96 H100 GPUs

(lmsys.org)
281 points GabrielBianconi | 1 comments | 29 Aug 25 14:07 UTC | HN request time: 0.25s | source
Show context
guerrilla ◴[30 Aug 25 09:42 UTC] No.45073311[source]▶
Now if only it would stop prefacing all its output with "Of course!" ;)

This is why I use DS though. I think its the only ethical option due to its efficiency. I think that outweighs all other considerations at this point.

replies(1): >>45075832 #
7thpower ◴[30 Aug 25 16:18 UTC] No.45075832[source]▶
The only ethical option? Please help me understand the argument here.
replies(1): >>45076220 #
1. guerrilla ◴[30 Aug 25 17:06 UTC] No.45076220[source]▶
Everything else uses more energy for both training and inference. Reducing the energy footprint is our highest priority in this domain. It outweighs the other considerations like it being Chinese, run by a hedge fund, etc. None of that matters if we destroy our ability to live on this planet. DeepSeek is not good enough, but we need to choose it in order to encourage competition on this front specifically. It's more important that companies focus on that than spend time improving other metrics.