To essentially solve the problem that the control performance of model predictive current control (MPCC) depends on the accuracy of model parameters, a model predictive current control of permanent-magnet synchronous motors (PMSM) based on the nonparametric prediction model (NPM-MPCC) is proposed. Firstly, the MPCC method based on conventional prediction model is introduced, and the model errors of the conventional prediction model on the control performance is analyzed. Then, a nonparametric prediction model for PMSM drive is proposed, which includes d-axis current prediction model and q-axis current prediction model. This proposed model doesn’t include motor parameters, and can achieve current prediction using prediction current error, sampling and stored information. Finally, experimental results verity the effectiveness of the proposed method.