Text model

VR-1

VR-1 is a frontier reasoning model trained specifically for cyber defense. Given a foothold inside an enterprise, it maps the environment and chains weaknesses across systems to find and verify real attack paths, then checks whether proposed fixes actually close the gap. On IntrusionBench it finds attack paths twice as effectively as other frontier models at one fourth the cost.