GitHub Copilot lets users choose cost, balance or intelligence

GitHub is adding three preferences to Copilot Auto: Efficiency, Balance and Intelligence. Auto still chooses a model for every prompt, but users can now indicate whether cost, speed or answer quality should dominate.

Quick answer

A preference, not a fixed model

All tiers use the same model catalog. Even under Intelligence, a simple docstring request may go to a smaller model. The tier changes the routing tradeoff; it does not guarantee one model or one price.

Billing follows the actual choice

Usage is charged according to the selected model. Paid subscribers retain a 10% Auto discount. Teams should observe consumption before applying Intelligence everywhere because maximum theoretical quality is unnecessary for routine edits.

Where it is rolling out

The setting is reaching VS Code, Copilot CLI and the GitHub Copilot app. Compare it with stable tasks, measure cost, delay and review quality, then choose a repository-appropriate default.

Toward explainable routing

Automatic routing becomes more useful when tradeoffs are visible. GitHub says more customization is coming. The next requirement is enough post-response detail for teams to understand what they paid for and why Auto changed models.