logoalt Hacker News

generalizationstoday at 4:50 PM0 repliesview on HN

Presumably a deepswe benchmark, which IIRC puts GLM 5.2 between opus 4.8 and fable.