Path 01 · Getting started

The curious one — « I just want to try it, hassle-free. »

You’ve heard about local LLMs and want to see what they can do on your machine. No code, no complicated command line—in 15 minutes, you’ll be chatting with your first model, offline and free.

Total duration

~15 min

Required level

None

Cost

0 €

Machine type

Recent PC or Mac

↑

Our promise

By the end of this walkthrough, Ollama will be running quietly in the background, you will have Mistral 7B downloaded locally, and a clean web interface for chatting with it, accessible from your browser. All in under fifteen minutes, without understanding a single line of code.

Who it's for this path is for you

We'd rather tell you early than let you waste 30 minutes for nothing.

✓ This is for you if

✓You barely know what an LLM is, but the topic intrigues you.
✓You want to test without entrusting your questions to an American service.
✓You are looking for the shortest path between “I’ve heard of it” and “it runs on my machine.”
✓You have a PC or Mac less than 5 years old with at least 8 GB of RAM.

✕ Look elsewhere if

·You want to fine-tune a model on your data: see the developer path.
·Want a professional setup with RAG on your documents? See the confidential professional track.
·Your machine dates from 2015 and has 4 GB of RAM: it will run, but it will crawl.

The path in 3 steps

Each step points to a detailed guide you can read alongside it. The order is optimized: don’t skip a step the first time.

Total · ~15 min

Recommended models

Three choices tested for this path—from lightest to most capable. Click for the detailed specifications (VRAM, context, benchmarks).

Getting started

Mistral 7B Instruct

Mistral AI · 7B · 5 GB in Q4

The default starting point. Native French, reasonable size.

View details →

Referenced model

qwen2.5-7b

Excellent quality-to-size ratio, very good at coding and reasoning.

Referenced model

llama-3.1-8b

Meta, multilingual, with longer and more structured responses.

Frequently asked questions

The real questions we get by email and on Mastodon. If yours is missing, open a ticket.

During a conversation: yes, your GPU runs at full load (200 to 400 W depending on the model). At idle between questions: almost nothing. For occasional use, this is negligible compared with a video game.

Once this process is complete

Next path

The confidential pro — « My files never leave my workstation. »

Juriste, médecin, RH, consultant, expert-comptable, notaire — vos documents sont sous secret professionnel ou NDA. Un LLM local couplé à un RAG vous laisse les interroger, résumer,…

→

The Local AI Kit

Your private ChatGPT, free, on your own machine in an hour — LM Studio, Ollama, Open WebUI, your documents, no cloud.