The local llama subreddit wiki has good info about RAM requirements: https://www.reddit.com/r/LocalLLaMA/wiki/models/