Welcome • User Guide
ToS • Privacy • Canary
Donate • Bugs • License

©2025 Poal.co

824

Anyone running this stuff? Just wondering. I don't know anyone with these running right now.

Archive: https://archive.today/FtDXp

From the post:

>I’m running a 24-billion parameter model on a Mac I bought in 2022, fully on-prem, fully under my control, powering an autonomous AI agent team through OpenClaw. The agents handle Slack, run research, draft content, monitor infrastructure, and spawn sub-agents for compound tasks. No API fees. No cloud provider. $1.50 a month in electricity. It took 14 errors, two hard freezes, and a week of patching OpenClaw’s source code to get here. This is everything I learned.

Anyone running this stuff? Just wondering. I don't know anyone with these running right now. Archive: https://archive.today/FtDXp From the post: >>I’m running a 24-billion parameter model on a Mac I bought in 2022, fully on-prem, fully under my control, powering an autonomous AI agent team through OpenClaw. The agents handle Slack, run research, draft content, monitor infrastructure, and spawn sub-agents for compound tasks. No API fees. No cloud provider. $1.50 a month in electricity. It took 14 errors, two hard freezes, and a week of patching OpenClaw’s source code to get here. This is everything I learned.
You are currently inside a comment thread.
Click here to see all the comments (14).
[–] • 1 pt

I’ve been watching videos about it, and I might give it a try.

What crApple Silicon CPU/GPU (cores) are you running it on? The second hand market is filled with cheap M1/2/3 with 16/24/32/48GB of Unified RAM. I might consider buying one to run a local instance and experiment with it.

[–] • 2 pts

Me, None, atm. I am working on some old AMD GPU's from various upgrades passed through in VM's.

I finally am working on this when I have time.

[–] • 1 pt

Get the max at worst, pro max at best. The M2 Max I have works any of it all relatively well. I have an isolated box for when and ollama. But have tested on th M2 Max on the pro book

[–] • 0 pt

I think 32/48GB will give decent results without swapping and wearing the SSD.

[–] • 0 pt

Apple M2 GPU cores 8–10 Apple M2 Pro GPU cores 30–38

As I see it, 3x at minimum

[–] • 1 pt

Also I been toying with buying a 3070 and dropping it in a 5,1 Mac Pro tower with dual Xeon. The machine has warping power and would run the 3070 fine. Then use that for some stable diffusion meme war battles.

[–] • 1 pt

I hope electricity bill is cheap where you live.

I was thinking about a second hand crApple M chip because they consume way less for equivalent results.

[–] • 1 pt

It cheap enough and I wouldn’t be using it constantly.