Flexible finger exoskeletons for hand assistance need multicontact forces sensing ability to guarantee good interaction between the exoskeletons and the human finger. However, force sensing of flexible fingers, especially in multicontact situations, remains challenging. To solve this problem, this article first proposes a variable-curvature quasi-static analytical modeling method for a novel nonparallel flexible finger, considering multipoint external forces and friction. Second, a force sensing method of the flexible finger is proposed based on the quasi-static analytical model to realize single-contact and multicontact force estimation. Third, a parameter optimization method is proposed to obtain optimal parameters of the quasi-static model to improve the force sensing accuracy. Finally, single-point and multipoint experiments of the flexible finger are conducted to verify the proposed force sensing method. The experimental results show that the median mean sum of the force sensing errors of the proposed force sensing method with the optimized parameters can be reduced by up to 75.9% compared with the force sensing method with the initial parameters.