GitHub Copilot is cutting the cost of each request by reusing cached context and routing tasks to the most efficient model, rather than always sending all information or using the most powerful option.

What happened: GitHub is rolling out two changes to Copilot in VS Code. Prompt caching lets Copilot reuse model state for repeated information instead of recomputing it on every turn. Tool search lets the model load tool definitions on demand rather than sending every tool definition into context each turn. Separately, Auto model selection is expanding to choose the right model based on task type—quick explanations, focused edits, and multi-file changes now get routed to different models rather than treating all tasks the same way.

Why it matters: Longer Copilot sessions require a lot of recurring information (instructions, context, conversation history, available tools, task state), and sending all of it repeatedly wastes tokens and costs money. By caching what repeats and routing tasks to models that fit the work, Copilot can deliver the same quality outcome with less unnecessary computation. The body notes that in evaluations, 'no single model consistently performed best across tasks'—a more efficient model often reaches the same result as a stronger one when the task does not require deep reasoning.

What to watch: Auto with task intent is already live in Visual Studio Code, github.com, and mobile. The rollout is expanding to Copilot CLI, GitHub App, and additional IDEs. Copilot Free and Student plans will be simplified to use Auto as the only model selection option, and admin controls will let organizations set Auto as the default or enforce it as the only option.