TestedliveAI Gateway

Gateway Router

Send a task through the routing rules and watch which model it lands on, and what that saves.

DYDDDylan and David

Eval labelCost and latency bench

Try it

The working version.

Mocked for this demo, with the numbers taken from the run that produced it.

Rules evaluate in order · first match wins

r1Task is conduct- or advice-bearing fired
Cost of being wrong is regulatory, not commercial. Never routed down.
r2Output feeds an agentic loop with tool calls
r3Structured extraction against a fixed schema
r4Classification into fewer than 20 labels
r5Anything else

Routed to

Frontier (reasoning)

direct API · p50 4200 ms

Customer requests excess waiver citing financial hardship…

Monthly cost at 18,000 calls

$319baseline

This route$319

Frontier for everything$319

Baseline is what most teams actually run: one strong model for every call.