Check Whether Picking an Ollama Model in Copilot CLI Turns On Offline Mode

From which version can I discover local models?

The GitHub announcement says that from CLI 1.0.94-0, /model can find the supported models of a running local Ollama. Models you have already configured and GitHub-provided cloud models appear alongside them. A model showing up in the list does not mean the session model has changed.

How do I put a discovered model into a session?

You pick a model, confirm the provider and endpoint address, and then either add and apply it to the current session, or only add it and leave your selection unchanged. The Ollama runtime and the model must be prepared first, and the official setup documentation lists tool calling and streaming as required.

So check in two steps. First, ask a short question to see whether a response comes back, then try a tool call such as reading a file in a separate throwaway task. A model that handles ordinary chat is not necessarily compatible with tool execution for development work.

How do a local selection and an offline setting differ?

The GitHub documentation was checked on 2026-10-10. If an offline environment is your goal, check not only the mode name but also whether the actual provider address is local.

What should I check when adopting this on a team?

Record the model's terms of use, the tool response format, and the device resources it uses. If you connect a company repository, the model address and access permissions should also match what your team allows. The intelligent local model routing mentioned in the announcement is guidance to wait longer for availability, so do not treat it as being in the same state as the discovery feature available now.