Self-hosting AI: what rigs do local LLM runners actually use?

Fun_Kangaroo512 · reddit · 2026-09-06

A Reddit thread asks self-hosters to share their local AI setups. The typical architecture described: a high-RAM desktop or server runs the model around the clock and is accessed remotely, so the client device can stay lightweight — even a MacBook Air or a smartphone.

The poster asks commenters to list full specs, making the thread a useful reference for anyone planning to build a local LLM stack (RAM/VRAM sizing, model choices, remote access patterns).

Original post →

More from Infra

Infra channel →