Guides for running AI locally — setup, hardware, and fixes.

Straight answers on local LLMs, the machines that run them, and OpenClaw — tested by hand, written plainly. No padding.

Read the local AI guide ↗ Or jump to the gear ladder
Today’s picks The three machines we point people at most. Prices move weekly — check before you buy.
BEST VALUE Used RTX 3090 · 24GB The default local-LLM card Check price on Amazon ↗ FASTEST 24GB RTX 4090 · 24GB Same fit class, much faster Check price on Amazon ↗ QUIET HOST Mac mini M4 · 24GB Leave-it-on agent box Check price on Amazon ↗
Get set up Install OpenClawA clean first install Your first 30 minutesWhat to do after install Pick a local modelBy RAM and use case
Fix a problem Troubleshooting hubGateway, auth, tunnel, VPS Slow local LLMEven when the model fits Out-of-memory crashesQuant, context, offload
Buy hardware Recommended gearThe GPU and Mac ladder Best GPU by budget3060 to RTX PRO 6000 Mac vs PC for local AIUnified memory vs VRAM

Reading list

Most useful guides

All 339 guides →
001

Best local LLMs for 128GB RAM

RAM guide 002

Best local LLMs for 64GB RAM

RAM guide 003

Best local LLMs for 48GB RAM

RAM guide 004

Best local LLM for RTX 4090

GPU guide 005

Best local LLMs for 32GB RAM

RAM guide 006

Best local LLM for RTX 3090

GPU guide 007

Best local LLMs for 96GB RAM

RAM guide 008

Best local LLM on the M4 Max

Mac guide 009

Best local LLM for RTX 5090

GPU guide

OpenClaw DC is an independent guide to local AI and OpenClaw, written by Saurav, a cloud consultant in Washington, DC. Some gear links are Amazon affiliate links.