Copilot routes tasks to right model automatically

Summary

Eliminates model selection friction in long agentic sessions and cuts token waste by matching task complexity to model capability. Prompt caching and deferred tool loading mean your context budget goes toward actual work, not repeated definitions.

Why it matters

Eliminates model selection friction in long agentic sessions and cuts token waste by matching task complexity to model capability. Prompt caching and deferred tool loading mean your context budget goes toward actual work, not repeated definitions.

Sources

1. 1.Prompt caching helps Copilot reuse model state for repeated prompt prefixes instead of recomputing the same prefix on every request.
2. 2.In our evaluations, no single model consistently performed best across tasks.
3. 3.Auto combines two signals: what model is healthy and available right now, and what kind of work Copilot is being asked to do.
4. 4.HyDRA (Agg.) balances quality for 72.5% savings
5. 5.routing accuracy stayed within four points of the English baseline across language groups, with no statistically significant quality gap
6. 6.Auto is the strong default for many tasks because it chooses a model based on what you are trying to do