How Copilot auto mode selects the best AI model | GitHub Checkout

Nhu explains GitHub Copilot’s new auto mode and how it routes prompts to the best available AI model based on reasoning needs and real-time capacity, with a quick demo in the Copilot CLI and VS Code.

Overview

In this GitHub Checkout episode, Nhu introduces auto mode for GitHub Copilot, which selects an AI model for a given prompt so developers don’t have to manually try different models.

Key points covered:

• What auto mode is: a Copilot mode that routes a prompt to the best available model.
• How selection works at a high level:
  • Routes based on the prompt’s reasoning needs.
  • Accounts for real-time capacity (availability) when choosing a model.
• Demo environments:
  • GitHub Copilot in the CLI
  • GitHub Copilot in VS Code
• How it was built (high level): discussion of the “intelligent routing” approach behind auto mode.
• Model evaluations and what’s next: how model evaluation informs routing decisions and future improvements.

Video chapters

• 00:00 — Welcome to GitHub Checkout
• 00:23 — What is auto mode?
• 01:21 — Demo: how auto selects models
• 04:23 — How we built intelligent routing
• 07:17 — Model evaluations and what is next