VerQen is a foundation decision model from ZorQelis AI: state in, decision out. Give it a policy, a case and options; it returns a decision with confidence and an escalation flag.
Not a prompted LLM - trained from scratch as a decision model. Open weights (Apache 2.0) AND open training data, so the release is checkable. Measured 8.4 ms per decision on an RTX 5060 Ti; $1.81 per million. Quickstart reproduces its worked example (deny_refund, P_correct 0.6656).
