aichannel

Running ollama on a m2 with 8gb of ram was a stupid idea. I got the following under-used machines that I'm wondering if I can take them apart or use to run local llm

- macbook pro 2012, 4gb ram but could upgrade. 2nd CPU slot got burned and no idea how to fix it
- macbook air 2018 that charges but doesn't turn on
- pi4 (currently used for a keyboard synthesizer)

These are some old machines save for the pi. MacBooks are also notoriously hard to disassemble (especially the air). 

I’ve used this - but not on machines that you have. Also maybe replacing the base os with Linux might help you. 

Ollama on the pi will work but will be slow…

TogetherCrew.com | single parent x 4 | seeker of adventure | survivor of chaos | host /togethercrew /kbc /single-parenting
seemore.tv/free/fc/kbc

Las Palmas de Gran Canaria, Las Palmas, Spain

Applied AI, databases, data, vibe coding (sic.), open source, community/devrel, remote. 

itinerant businessperson. globalist.

Your machines are far too old for regular macOS. 

But ask yourself: what is the purpose of running a local LLM?

Your m2/8gb ram could run a small mistral pretty easily I reckon. Slow token output, but should be “ok”.

I'll need to decide if this is just a nice side project to try out something new. And if yes, what will I learn. someone else already planted the idea in my mind to replace the os with linux

These are some old machines save for the pi. MacBooks are also notoriously hard to disassemble (especially the air). 

I’ve used this - but not on machines that you have. Also maybe replacing the base os with Linux might help you. 

Ollama on the pi will work but will be slow…

https://github.com/exo-explore/exo

You need more RAM and a better CPU. Try with Gemma LLM as it might work.

i think your only reasonable bet is gemma 2b personally everything else prob too slow

as noted below you can also try exo