GitHub Copilot CLI: Auto Model Selection Now Generally Available
GitHub CopilotGitHub has made auto model selection generally available in GitHub Copilot CLI, completing its rollout across all major Copilot surfaces. With auto enabled, the CLI dynamically routes each session to the best-performing model — including GPT-5.4, GPT-5.3-Codex, Claude Sonnet 4.6, and Haiku 4.5 — based on availability, rate limits, and administrator policies. All paid subscribers receive a 10% discount on model multipliers when using auto, and users retain full visibility into which model was selected for each session directly within the CLI interface.
Sources & Mentions
4 external resources covering this update
Auto Model Selection Arrives in GitHub Copilot CLI
GitHub has extended auto model selection to GitHub Copilot CLI, making the feature generally available for all paid Copilot plans. Previously, CLI users had to manually choose from a growing list of AI models — a friction point that required awareness of each model's strengths and cost profile. Auto mode removes that overhead by delegating the choice to Copilot itself.
How Auto Model Selection Works
When a developer selects auto as their active model in the CLI, Copilot dynamically routes each session to the most efficient model available at that moment. The pool currently includes GPT-5.4, GPT-5.3-Codex, Claude Sonnet 4.6, and Haiku 4.5, with the set expected to grow and evolve over time as GitHub adds and retires models.
The routing logic accounts for two key factors: rate limits and model availability. If a preferred model is under heavy load or approaching rate constraints, auto quietly shifts the session to an equally capable alternative — providing continuity without manual intervention.
Transparency and Control
Auto mode is opt-in and non-permanent. Developers can switch back to a specific model at any time using the CLI's model picker. After each session, the CLI surfaces which model was selected, giving users full transparency into what ran on their behalf. This visibility matters especially for developers who need to audit model usage or understand cost attribution across sessions.
Administrator model policies are also honored: if an organization has restricted access to certain models, auto will not route to those models regardless of availability or performance.
Billing and Discounts
Premium request billing under auto mode is based on whichever model Copilot selects. Currently, auto only routes to models with multipliers between 0x and 1x, keeping costs predictable. On top of this, all paid Copilot subscribers receive a 10% discount on model multipliers when using auto. For example, a model with a standard 1x multiplier draws only 0.9 premium requests per use — a modest but consistent saving for heavy CLI users.
Availability
Auto model selection in the CLI is generally available as of April 17, 2026, included in Copilot CLI v1.0.32-0. It is available to all paid Copilot subscribers — Pro, Pro+, Business, and Enterprise — with no additional configuration required beyond selecting auto in the model picker.