This paper introduces a semiparametric method of estimating multinomial models that imposes extremely weak monotonicity assumptions about a function of observable characteristics. Previous methods have imposed stronger, typically parametric, conditions on this function. The only assumptions made in this paper about the function of characteristics are its monotonicity, upper-semicontinuity, and uniform boundedness. The method is applicable, among others, to polychotomous choice models. The estimation method is shown to be strongly consistent. A technique to calculate the estimator is provided.