Get Started
← Own your AI · Mission 12 of 31

Cancel Copilot: a free, private coding assistant in VS Code on your own laptop

No subscription, no usage limits, no code sent to anyone. Continue puts chat, one-instruction edits and grey autocomplete into VS Code, powered by two small models running on your own laptop.

30 minIntermediateTools: FreeStep 1 free
Tested 9 Oct 2026 · Continue 2.0.0, VS Code 1.136.1, Ollama 0.40.1, Gemma 4 e2b, Qwen2.5-Coder 1.5B, Windows 11 (16GB, no graphics card)
A chat panel in VS Code that can read your files and explain them
Select code, press Ctrl+I, describe the change, and get a red and green diff to accept or reject
Grey autocomplete suggestions as you type, accepted with Tab
  • Real timings from an ordinary laptop with no graphics card, and the one setting that stops suggestions being cut off halfway
  • An upgrade path to a stronger coding model when you have the hardware for it
Before you start
  • Ollama with gemma4:e2b downloaded. Not done that yet? Do Run a ChatGPT-style AI on your own laptop first (15 minutes). Gemma 4 needs Ollama 0.20.7 or newer: check with ollama --version.
  • VS Code (free), and a folder of code to try it on. Any small project works.
  • About 1GB of disk space for the autocomplete model and 230MB for Continue
  • 16GB of RAM recommended: you'll have two small models loaded at once
01/06Download the two local models0 done
STEP 01 of 06~4 min

Download the two local models

One model for chatting and editing, a tiny fast one for autocomplete. Both run in Ollama on your machine.

  1. If you've done the Ollama lesson you already have Ollama and Gemma 4. If not, install Ollama first:

    Windows (skip if you have Ollama)
    winget install --id Ollama.Ollama -e
    Mac / Linux (or download the app from ollama.com/download)
    curl -fsSL https://ollama.com/install.sh | sh

  2. Now download the chat model (skip if you have it) and a small model made for code completion. Qwen2.5-Coder 1.5B is the one Continue's own docs recommend for local autocomplete:

    Terminal
    ollama pull gemma4:e2b
ollama pull qwen2.5-coder:1.5b

  3. The coder model is 986MB, about a minute on a normal home connection. Check both are there:

    Terminal
    ollama list
    Output
    NAME                  ID              SIZE      MODIFIED
qwen2.5-coder:1.5b    d7372fd82851    986 MB    Less than a second ago
gemma4:e2b            b37049369adf    4.6 GB    About an hour ago
    TipWhy two models? Autocomplete fires every time you pause typing, so it needs to be tiny and quick. Chat and edits need a model that's better at following instructions, and can afford to take longer.
    TipGot a graphics card or 32GB of RAM? Continue's docs list Qwen2.5-Coder 7B as the stronger open model for autocomplete. Run ollama pull qwen2.5-coder:7b (4.7GB), then in step 3 change qwen2.5-coder:1.5b to qwen2.5-coder:7b in config.yaml. On a laptop with no graphics card, start with 1.5B: the 7B model is several times bigger and noticeably slower to suggest.
✓
Checkpointollama list shows both qwen2.5-coder:1.5b and gemma4:e2b.
If it breaks'could not connect to a running Ollama instance'
'could not connect to a running Ollama instance'
Ollama isn't running in the background. Open Ollama from the Start menu (or Applications on a Mac) and try again.
Still stuck? Ask the community
Next mission · Own your AIAn AI pair programmer in your terminal that edits your code and commits it, using a free local model
→