Skip to content

Comment on How to run DeepSeek R1 locallyparent

Comments

“you can get a dual EPYC server with 768GB RAM - CPU inference only at around 6-8 tokens/sec.”

This is what I run at home. I built it just over a year ago and have run every single model that has been released.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.