GitHub has added local Ollama model discovery to GitHub Copilot CLI version 1.0.94-0, letting developers use /model to find supported local models alongside configured and Copilot cloud models. The change lets a compatible discovered model be used in the current CLI session without a restart.
Discovering local choices
The command looks for models from a running local Ollama instance. Discovery alone does not add a model: users inspect its provider and endpoint, then choose either to add and use it for the session or to add it without changing the active model.
Requirements for a model
Ollama and the desired model must already be installed. This workflow does not install the runtime or download a model, and eligible models must support both tool calling and streaming.
Connection feedback
If Copilot CLI cannot connect to a provider, the model picker displays an explanation intended to identify the connection issue that needs attention.
Local is not automatically offline
Selecting a local model does not enable offline mode or switch off GitHub telemetry. GitHub says offline mode remains an explicit COPILOT_OFFLINE=true choice, and a remote provider may still receive prompts and code context over the network even when that setting is enabled.
Routing remains pending
GitHub also said it is announcing intelligent routing with local models, while asking users to stay tuned for availability.
Definition. Local Ollama model discovery is a Copilot CLI feature that finds compatible models from a running local Ollama instance.
| Action | Result |
|---|---|
| Add and use a discovered model | Adds the model and changes the active model for the current session. |
| Add without changing the active model | Adds the model without changing the current active model. |
Key takeaways
- Ollama and the desired model must already be installed.
- Eligible models must support tool calling and streaming.
- Discovery does not add a model automatically; users can add it with or without switching the active model.
- The model picker explains connection issues when Copilot CLI cannot reach a provider.
- Selecting a local model does not automatically enable offline mode or disable GitHub telemetry.
- Offline mode requires setting COPILOT_OFFLINE=true, and remote providers may still receive prompts and code context.
FAQ
How do developers find local Ollama models in Copilot CLI?
They use the /model command while a local Ollama instance is running.
Does Copilot CLI install Ollama or download models?
No. Ollama and the desired model must already be installed.
What makes a local model eligible?
It must support both tool calling and streaming.
Does choosing a local model enable offline mode?
No. Offline mode remains an explicit COPILOT_OFFLINE=true choice.
