The Continuity Layer
Run your own AI · Updates and new versions

Updates and new versions

thecontinuitylayer.com/open/learn/updates

Checked 6 November 2026. US dollars. No affiliate links.

Open models move fast: a better one of the same size lands every few months. The good news is that on your own machine nothing changes until you decide. Here is how to update on your terms.

By the end

A newer model running alongside the one you trust, your settings intact, and your disk tidy.

Update the app

  • Ollama on Mac and Windows downloads updates itself; choose "Restart to update" from its menu bar or taskbar icon. On Linux, run the install script again.
  • LM Studio offers updates inside the app.
  • Open WebUI in Docker: pull the new image and recreate the container. Your chats live in its data volume, so they stay.

Update a model

ollama pull gemma4:12b     # fetch the newest build of that tag
ollama list                # what you have, with sizes
ollama rm gemma4:e4b       # remove one you no longer use

Pulling a tag you already have downloads only what changed. To try a new family, pull it next to the old one and compare the two on the same questions before you switch.

Where the models live

SystemOllama's model folder
macOS~/.ollama/models
WindowsC:\Users\<you>\.ollama\models
Linux (installer)/usr/share/ollama/.ollama/models

To keep models on a bigger drive, set OLLAMA_MODELS to another folder. From Ollama's FAQ, read on 6 November 2026.

Three habits

  • Keep your instructions in a file, not just in an app. A Modelfile or a context file moves to a new model by changing one line.
  • Read the license of a new version before using it for work. Licenses change between versions.
  • Keep one model you trust installed until the new one has earned its place.
← Chat with your own documentsNext: A local AI server for the house →