LLM Router

An LLM router is a component that directs each incoming request to the most suitable language model, balancing quality, cost, and latency across a set of available models.

A router might send simple queries to a small, cheap model and hard ones to a larger model, lowering average inference cost without sacrificing quality where it matters.

Routing is a practical lever on the cost of AI labor, and its decisions are best tuned against measured task outcomes.