GitHub Copilot CLI discovers local Ollama models
Copilot CLI 1.0.94-0 discovers eligible Ollama models through /model; installation, offline mode and telemetry settings remain separate.
GitHub Copilot CLI can show eligible Ollama models in its picker and asks the user to confirm adding them. It does not automatically enable offline mode.
The CLI did not have the newly documented picker-based discovery flow for a running local Ollama instance.
Starting with CLI 1.0.94-0, /model lists supported running Ollama models; users explicitly approve adding them.
GitHub published an updated model-selection flow for GitHub Copilot CLI on October 7, 2026. From CLI version 1.0.94-0, the /model command can discover supported models exposed by a running local Ollama instance. Discovered models appear in the same picker as configured providers and GitHub Copilot cloud models, reducing the need to leave the CLI to find a usable local option.
Discovery is not automatic installation. GitHub states that Ollama and the model must already be installed; the picker does not install a runtime or download weights. Users review the provider and endpoint after discovering an eligible model and explicitly choose whether to add it and use it for the current session or add it without switching. A CLI restart is not required after that selection.
Only models that support both tool calling and streaming are eligible. This distinction matters for an agentic tool: a model needs to select actions, receive tool results and continue work, not simply generate text. GitHub also says provider connection failures surface in the model picker with an explanation so users can diagnose configuration problems.
Local inference and offline operation are not interchangeable claims. Selecting an Ollama model does not automatically enable offline mode or disable GitHub telemetry. Copilot CLI still requires an explicit COPILOT_OFFLINE=true setting for offline mode. Using a remote provider can transmit prompts or code context over the network even when offline mode is configured, depending on the provider connection. Teams should inspect the provider and offline documentation before making data-handling claims.
This update is scoped to GitHub Copilot CLI, not to the separate cloud coding agent. It gives developers using local models a more convenient discovery and selection workflow but does not establish universal guarantees about data residency, privacy compliance or network isolation. Those questions still depend on the selected endpoint, the runtime and the surrounding enterprise configuration.
Primary source
Sources behind this update
AgentenCode publishes product changes only after source-first review. The original provider source remains authoritative for current details.
Full context in the evidence profile.
Full context in the evidence profile.
Full context in the evidence profile.