This document presents research using genetic programming to develop non-intrusive models for estimating voice over IP (VoIP) call quality. The researchers used a VoIP simulation environment to generate distorted speech files under different network conditions. Genetic programming experiments were performed to evolve models relating mean opinion score (MOS) estimates to transport layer metrics like packet loss rate and jitter. The best models achieved mean squared errors within 0.04 of the intrusive PESQ algorithm estimates, performing significantly better than the ITU standard P.563 non-intrusive model. The models provide a simple, real-time alternative to PESQ for VoIP quality monitoring.