How do you run an LLM locally?

You run an LLM locally by installing a free tool like Ollama, LM Studio, or llama.cpp, downloading an open-weight model, and running it entirely on your own machine — no cloud, no API key, no per-token bill. The three tools cover different comfort levels:

• Ollama — one terminal command downloads and runs a model. The default choice for most people.
• LM Studio — a desktop app with a chat window and a built-in model browser. Zero terminal required.
• llama.cpp — the low-level engine the other two build on, for people who want maximum control.

The basic workflow is the same everywhere: pick a tool, browse its model library, download a model that fits your machine, and start chatting.

Memory is the whole game. The model has to fit in your GPU's VRAM or your Mac's unified memory, which is why Apple Silicon Macs with lots of RAM became the unofficial local-LLM machine. Quantized models — compressed versions that trade a little quality for a much smaller footprint — are what make this practical on consumer hardware.

The honest trade-off: a local model won't match Claude or GPT-5.6 on hard tasks. What you get instead is total privacy, offline access, and zero marginal cost — which for drafting, summarizing, and tinkering is often plenty.