What is Model Routing?

Model routing (or smart routing) is the automatic selection of which AI model or provider should handle a given request, based on task requirements, cost, policy, and performance. It decouples what employees ask from which model answers.

Not every task needs the most expensive frontier model. Routing sends simple tasks to cheaper models, sensitive tasks to approved or on-prem models, and regulated workloads away from prohibited providers.

Routing is also a governance control: it enforces which providers may see which data classes, and it removes single-provider dependency.

Tokto applies routing at the checkpoint as part of FinOps and policy enforcement.