AI Dynamics

Global AI News Aggregator

About

Llama 3-70B Performance: MMLU Benchmarks and GPT-5 Implications

Llama 3-70B is pretty close to what you would expect. Only GPT-5 (or an unexpected leap from somewhere else) will tell us about the future curve. (And also MMLU is super limited, but it is one of the few yardsticks we have)

→ View original post on X — @emollick