from Hacker News

[broken] Cursor trains a 70B model that achieves >1000 token/s using...

by jnnnthnn on 5/17/24, 2:53 PM with 1 comments