GitHub Copilot CLI Adds Local Ollama Model Discovery
Copilot CLI can now find compatible models from a running Ollama instance, but local selection does not automatically enable offline operation or disable telemetry.
Edited by Tyronne Panaino
GitHub announced on October 7 that GitHub Copilot CLI version 1.0.94-0 can discover supported models from a running local Ollama instance. The models appear in the same selection flow as models a user has already configured and cloud models provided through Copilot.
The update gives developers a direct way to inspect and select compatible local models without leaving the CLI. It does not install Ollama, download a model or automatically add anything it finds. Ollama and the model must already be installed, and GitHub says eligible models must support tool calling and streaming.
A model picker, not a runtime installer
The discovery flow presents a model's provider and endpoint for review before it is added. A developer can choose to add and use the model for the current session or add it without immediately switching. GitHub says the selected model can be used in the current session without restarting Copilot CLI.
Connection problems are also surfaced in the picker with an explanation. That makes discovery useful as a visibility layer as well as a selection mechanism, although the announcement does not provide a complete compatibility list for Ollama-served models. The stated requirements are tool calling and streaming support.
Local selection is not an offline guarantee
GitHub draws an important boundary around the word local. Selecting a model discovered through Ollama does not itself activate Copilot CLI's offline mode, and it does not disable GitHub telemetry. Offline operation remains a separate, explicit choice.
The source also warns that a remote provider can still receive prompts and code context over the network even when offline mode is enabled. For teams evaluating local models for privacy, compliance or network isolation, the relevant question is therefore the full provider and endpoint path, not only where the selected model appears to run. The fetched announcement does not enumerate telemetry fields, retention behavior or every possible provider configuration.
What changed and what remains open
GitHub describes the CLI update as building on the model-provider experience already available in the Copilot app. The new checkpoint is discovery inside Copilot CLI from version 1.0.94-0, followed by explicit review and session-level selection.
GitHub also announced intelligent routing with local models, but the same source says availability is still to come. That routing should not be treated as a shipped capability yet. The announcement provides no independent quality comparison, benchmark, latency distribution or reliability study for discovered models, so this article does not infer a performance or privacy advantage from local selection alone.
The next verifiable checkpoints are broader compatibility documentation, the actual availability of intelligent routing, and evidence describing telemetry and network behavior across supported configurations.
Status
Confirmed. Availability and scope come from GitHub's official changelog. Internal confidence is medium because the release is documented by one first-party source and no independent operational testing was fetched.
Sources
Update note: Last reviewed 2026-10-08. We will revise this post when GitHub documents material changes to model compatibility, intelligent routing, telemetry or offline behavior.
Sources
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.