GitHub Copilot Lets Users Pick Efficiency, Balance, or Intelligence Tiers for Auto Model Selection
GitHub Copilot’s auto model selection feature now offers three configurable tiers, letting users tune how it balances cost and quality when picking a model for each prompt.
Details
- Efficiency: prioritizes keeping costs low, suited to fast, straightforward tasks
- Balance: weighs cost, quality, and latency together, aimed at everyday work
- Intelligence: prioritizes quality, built for complex tasks
- How selection works: auto selection evaluates each prompt individually and picks an appropriate model from the same shared pool regardless of tier — even in intelligence mode, simpler requests may still route to smaller, more efficient models; charges are based on the model actually selected, not the tier chosen
- Pricing: paid subscribers get a consistent 10% discount on auto-selected model usage across all three tiers
- Availability: rolling out now in Visual Studio Code, Copilot CLI, and the GitHub Copilot app
What happened next
Giving users direct control over the cost/quality tradeoff in auto-selection, rather than a single fixed behavior, lets teams tune Copilot for cheaper, faster iteration on routine work while reserving higher-quality routing for harder tasks — with the discount held constant regardless of which tier is chosen.