Serrated milling tools are widely used for chatter suppression in roughing difficult-to-cut Titanium and Nickel alloys in the aerospace industry. Due to the complexity of chip generation and serration wave geometries ground on the flutes, the chatter stability diagrams are predicted with time marching numerical simulation or semi-discrete time-domain methods, which are computationally too costly to use in practice. This paper presents a frequency domain model of milling dynamics with variable delays caused by the flute serrations. The endmill is divided into discrete cylindrical elements, each having a different radius from the cutter axis. As the cutter rotates and cuts metal, the angular distance between the subsequent tooth varies as a function of serration amplitudes and feedrate; hence, the regenerative delays vary. The angular delays and effective directional factors are averaged for each tooth to form a time-independent but serration-dependent characteristics equation for all discrete cutter elements. The stability of the resulting characteristic equation of the system is solved using Nyquist theory and compared against the experimental results and existing time marching and semi-discrete time-domain solutions. The proposed analytical model predicts the stability charts about 30 times faster than the time-domain models while providing acceptable accuracy.