GitHub Copilot CLI Adds Discovery for Local Ollama Models

GitHub Copilot CLI added the ability to discover and select local models from running Ollama instances alongside configured cloud models.

Details

• How it works: users run the /model command to browse available models; the picker shows each model’s provider and endpoint, and after selecting a local model, users can “add and use for this session” or “add without switching”
• Requirements: the feature doesn’t automatically install runtimes or download models — Ollama and the models must already be installed, and models must support tool calling and streaming
• Error handling: provider connection failures appear in the picker with explanatory messages
• Not an offline switch: selecting a local model doesn’t enable offline mode or disable GitHub telemetry; offline mode still requires explicitly setting COPILOT_OFFLINE=true
• Availability: live now in CLI version 1.0.94-0 and later; GitHub also previewed upcoming intelligent routing between local and cloud models

What happened next

This gives Copilot CLI users a lighter-weight way to mix self-hosted models into their workflow without switching tools, landing alongside this same week’s local sandboxing GA and purpose-built secret-detection model as part of a broader push toward more developer control over how and where Copilot’s agentic features run.