Hacker News Reader

From my experience with complex coding tasks (AI infra), I don't think these open weight models are close.

Not sure if I agree, I tried GLM5.3 and it was pretty decent. Ok, it's not Opus, but maybe it's Sonnet?