Redson Dev · Idea
Run your own AI chatbot, fully offline, on your laptop
Published April 30, 2026
No API key, no monthly bill, no data leaving your machine. With Ollama and a 4–8B parameter model, even a 5-year-old laptop can run a useful assistant for code, writing and Q&A.
What you'll need
- Any laptop with 8GB+ RAM (16GB is comfortable)
- 10GB free disk space
- Internet connection (for the initial download only)
Step-by-step
- 01
Install Ollama
Go to ollama.com and grab the installer for your OS. It runs as a tiny background service.
- 02
Pull a model
Open a terminal: `ollama pull llama3.2:3b` for a small fast one, or `ollama pull qwen2.5:7b` if you have 16GB RAM. The download is 2–5GB.
- 03
Chat from the terminal
`ollama run llama3.2:3b` drops you into a chat. Type questions. Type `/bye` to leave.
- 04
Get a nicer UI
Install 'Open WebUI' (Docker) or use the free 'Msty' desktop app — both auto-detect Ollama and give you a ChatGPT-style interface.
- 05
Try a real task
Paste a CSV header and ask it to write a SQL query. Paste an error log and ask what's wrong. The 3B model is dumber than GPT-4 but faster than you think — and it's yours.
Tips
- Smaller models (1–3B) are great for autocomplete and summaries; reach for 7B+ for reasoning.
- Quantized models (Q4) are 4× smaller with barely any quality loss — Ollama uses them by default.
