Cogent Security releases VR-1.1 vulnerability repair model

Cogent Security released VR-1.1, a reasoning model that writes fixes for newly disclosed software vulnerabilities. Cogent says it outperformed Claude Code and Codex on vulnerability repair in its own evaluation. In that test, Cogent blocked outbound network access during repairs and hid the grading tests until each agent finished. Inference budgets differed across the three systems, according to the methodology Cogent Research published. Where the environment allows, VR-1.1 first writes its own proof of concept that triggers the reported flaw. Cogent sells AI agents that detect and fix vulnerabilities, and customers choose whether a human signs off on each change.

AI-generated from reports · 2 hours agoUpdate

Not enough continuous observation data yet; trend chart is not drawn for now.