Inference routing
Local-versus-cloud language routing: coding stays local, complex work proxies out.
Inference is routed by the kind of work: coding and routine tasks run against a local model, while complex requests proxy out to a cloud provider through the cost gate on your own account. This routing runs in production.
Local ⇄ cloud language routingLive
Cost gate (your own AI accounts)Live
Tip
Because spend only flows through the cost gate, local routing costs you nothing — only proxied requests touch your provider key.
Was this page helpful?
Last updated June 3, 2026