The craziest LLaMA 3 reveal: The 400B+ version of the model is **on par with Claude 3 Opus**, and it's still training. Soon, we'll have a better-than-Opus, fully open-source model. The implications are huge.

→ View original post on X — @mattshumer_