Start with local AI

A practical first path for beginners: size the hardware, choose one runtime, and keep uncertainty visible.

This start path avoids “best” claims. It gives conservative steps and links to records that show their verification status.

Begin with fit and privacy

This page should get a beginner from confusion to a safe first path without overselling what local AI can do.

Start with your hardware, not the tool. Local AI depends on RAM, VRAM, Mac unified memory, model size, quantization, context length, and whether your workflow stays local. Once you know your hardware tier, choose one first app, run one small model, and only then move into larger models, PDF chat, Open WebUI, or private workflows.

Pick a first path

Path A: Easiest desktop app

Use LM Studio if you want a graphical app for downloading models and chatting locally. Read Install LM Studio.

Path B: Best local runtime path

Use Ollama if you want a lightweight runtime, local API, or a base for Open WebUI later. Read Install Ollama.

Path C: Browser-style local workspace

Install Ollama first, then add Open WebUI. Read Open WebUI with Ollama.

Path D: Private document testing

Start with the privacy guide, then read Chat With PDFs Locally. Do not upload confidential documents until you know where the model, embeddings, files, and logs live.

Avoid these first-session traps

Do not download the biggest model first.
Do not assume “local app” means “local model.”
Do not assume 8GB system RAM is the same as 8GB GPU VRAM.
Do not expose a local server to your network until you understand the security model.
Do not use sensitive documents for the first test run.

Begin with fit and privacy

1. Check hardware fit

2. Pick a beginner runtime

3. Read the privacy guide