Is 32GB RAM enough for running local LLMs?

Yes, 32GB is the sweet spot — it can run 7B parameter models comfortably with 4-bit quantization and handle 13B models if you’re careful. You’ll need 4-5GB for a 7B model and 8-9GB for a 13B, plus 8-16GB for your OS and context. Don’t expect fast speeds without a beefy GPU, but it’s practical for experimentation and casual use.