A practical guide to running Qwen3 8B locally on a MacBook Air M4 using Ollama. Covers installation without Homebrew, PATH setup, starting the server, pulling the model, and three interaction modes: interactive chat, one-shot terminal commands, and HTTP API via Python. Also covers disabling chain-of-thought thinking tokens, a privacy caveat about Ollama's web search feature, and integrating the local model with VS Code via the Continue.dev extension for offline coding assistance.
Table of contents
Why Do This?The Machine and the SpecsWhy Ollama?Fine-Tuning the Experience — Taming the “Thinking” TokensA Warning About Web SearchBonus: VS Code IntegrationWhat if I have a Windows Computer?Where This Leaves Me67 Impressions