Dynamic modeling of an input-output coupled piezoelectric fast steering mirror(FSM) was discussed, especially in case that the sampling frequency of the sensor is so high (e.g., high-speed camera or position sensitive device is used) that the dynamics of the FSM is significant and can′t be neglected. The dynamics and input-output coupling in the FSM was analyzed. A method was proposed to model the dynamics and input-output coupling in the FSM, based on the input-output data and the subspace identification algorithm. Experiments were carried out for data collection and validation of the proposed method. The results show that the variance-account-for(VAF) of the dynamic input-output coupled model is as high as 95%, indicating high accuracy of the model. Comparison was made between the proposed model and static model as well, showing significant improvement on the model accuracy. The result can be applied in optimal feedback controller design for closed-loop FSM systems, to improve the closed-loop performance in laser beam drifting correction.