OmniRail routes agent requests by Jev difficulty score
OmniRail uses Jev to score each agent request on a custom difficulty scale and route it to a cheap or frontier model, all on one key and balance, at a fraction of a cent per decision.

Most questions an agent sends are easy. A few are hard. Sending all of them to the most expensive model wastes money; sending all of them to the cheapest one fails on the hard ones. So ask first. Jev scores each request on a scale you write (simple lookup, normal task, hard reasoning) and your code sends it to a cheap model or a frontier one. All of them are on the same OmniRail key and the same balance. The question to Jev costs a fraction of a cent. http://omnirail.org/models