What is the ram requirement for Mac to properly run it?
It is now (somewhat) possible to run fully local (no cloud, no spyware) Alexa/assistant alternative with Home Assistant, Ollama (I run it on Apple Silicon) and as an endpoint Home Assistant Voice Preview Edition.
I've written about my setup (and challenges), but TLDR: You can just say "Ok Nabu, turn on my bedroom AC" and it works.
I've dreamt about this in Cypherpunk Visions 2023-2025 (a short book). Now I am going to write update for 2026-2029 (3 years is a good timeframe) and this will be "normal" for those who want it now, it's not just a dream.
https://community.home-assistant.io/t/blueprint-on-ai-using-ollama-on-apple-silicon/916158
Discussion
Not much, 2gb for smaller models, around 8GB for llama. Then some more for whisper (500MB maybe)
I run it on 32GB mac mini.