The in-plane longitudinal compressive strength of continuous fiber-reinforced composites is significantly less than their tensile strength, a phenomenon that is still not well understood. Dow and Rosen attempted in 1965 to explain this phenomenon through a fiber microbuckling model, in which each fiber was modeled as a beam-on-an-elastic-foundation. The model, however, grossly overestimates the compressive strength; researchers have since focused on introducing factors that reduce the predicted compressive strength to fit experimentally observed values. These include initial fiber waviness, fiber and matrix properties, the interfacial strength, and interphase properties. While empirical results have demonstrated the influences of some of these factors, an analytical model that can accurately predict the strength while taking into account the underlying failure mechanism is still lacking. The present study treats the compressive strength prediction of continuous fiber-reinforced composites using a different approach. Here, the matrix surrounding the fiber is not assumed to be homogeneous, but is divided into a discrete interphase region adjacent to the fiber and bulk matrix phase beyond the interphase. The three-phase model is shown to provide a more accurate prediction of the compressive strength of typical high-performance composites, such as glass and carbon fiber-reinforced epoxy. The present model's prediction agrees well with the trends shown in the experimental database available in the literature.