Travel time is very critical for emergency vehicle (EV) service and operations. The characteristics of EVs travel time are quite different from those of ordinary vehicles (OV). Although EV owns highest road privilege, it may still experience unexpected delay that results in more costs to the society. In this study, we model such delay as extreme conditions. Generalized extreme value (GEV) theory and distributions are employed to measure and analyze this extremity. With maximum likelihood estimation (MLE), Frechet and Weibull distributions are applied to model the unit travel time of EVs. Kolmogorov–Smirnov test shows that both distributions fit very well. On the basis of GEV theory, two studies are conducted respectively. First, for large sample observations, an extremity index is developed to measure EVs’ travel time variability. Second, for a single observation, a regression model is proposed to identify potential significant influential factors to impact EVs link travel time. It is found that distance is one of the major influential factors. Longer distance will result in less unit travel time. Moreover, the occupancy and time of day also affect the travel time of EVs in different road conditions.