The invention provides a hybrid computing method for dynamic modeling of a foot-type robot. According to pre-stored basic physical parameters of the foot-type robot, angles and angular speeds of eachjoint of the foot-type robot are sampled, and sampling values fed back by sampling are acquired; according to the sampling values, a transfer matrix of coordinate rotation is worked out; according tothe basic physical parameters and a computing method for an inertia matrix based on a Lagrange method, the inertia matrix of each connecting rod is deduced, and an inertia matrix of the foot-type robot is obtained through matrix summation; and according to the sampling values and the basic physical parameters, improvement is conduced based on a Newton-Euler method, and relevant vectors of centrifugal force, Coriolis force and gravity of the foot-type robot are obtained through iterative computation, so the dynamic modeling of the foot-type robot is completed. Through integration of computationadvantages of the Lagrange method and the Newton-Euler method and amendment at different stages, computation efficiency of the dynamic modeling is increased.