Guides
Learn to run AI on your own hardware
Clear explanations of the things that actually decide what you can run, and how fast.
Explainer
How much VRAM do you need to run LLMs locally?
Weights, KV cache and overhead, worked out with real numbers for the open models people actually run in 2026.
Updated Oct 14, 2026
Explainer
LLM quantization explained: what Q4_K_M, Q8_0 and the rest mean
The letters and numbers decoded, what the measurements actually show, and a simple way to choose a quant for your GPU.
Updated Oct 14, 2026
Buying guide
The best Macs and mini PCs for running LLMs locally (2026)
Unified-memory machines hold models that no affordable graphics card can, but their slower memory means you have to choose your models with care.
Updated Oct 14, 2026
Buying guide
Buying guides
Buying guide
The best GPUs for running AI models locally (2026)
Memory decides which models you can run, and memory bandwidth decides how fast they answer. These are the cards that give you the most of both for the money.
Updated Oct 14, 2026
Buying guide
The best Macs and mini PCs for running LLMs locally (2026)
Unified-memory machines hold models that no affordable graphics card can, but their slower memory means you have to choose your models with care.
Updated Oct 14, 2026
Explainer
Explainers
Explainer
How much VRAM do you need to run LLMs locally?
Weights, KV cache and overhead, worked out with real numbers for the open models people actually run in 2026.
Updated Oct 14, 2026
Explainer
LLM quantization explained: what Q4_K_M, Q8_0 and the rest mean
The letters and numbers decoded, what the measurements actually show, and a simple way to choose a quant for your GPU.
Updated Oct 14, 2026
How-to
How-tos
Opinion