Sags are freeway sections along which gradient changes significantly from downwards to upwards. Sags are often bottlenecks in freeway networks. Capacity decreases due to changes in car-following behavior. We present a microscopic traffic flow model that describes driving behavior and traffic flow dynamics at sags. The model includes a new longitudinal driving behavior model that accounts for the fact that drivers generally have a limited ability to compensate for the negative effect that an increase in slope has on vehicle acceleration. The face-validity of the microscopic model is tested by means of a simulation study. The study site is a sag of the Tomei Expressway (near Tokyo). The simulation results are compared to empirical findings presented in the literature. We show that the model is capable of reproducing all the relevant phenomena that cause the formation of congestion at sags.