1 Comment

BLOOMChat and the OpenLLM Leaderboard show us a simple trend in LLMs right now: size still matters. As data improves, I suspect this matters less.

LLaMA 65B is top still: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard, it may change as we add more eval types though.

Expand full comment