Parameter mismatches in conventional model predictive current control (MPCC) would worsen torque and THD performance. To eliminate the parameter dependence of MPCC, this letter proposes a model-free predictive current control (MFPCC) for a three-level inverter-fed interior permanent magnet synchronous motor (IPMSM). Firstly, to increase the current difference updating frequency in the diode neutral-point-clamped (NPC) three-level inverter, a synchronous updating method is proposed in a manner where the number of updated current differences during one sampling period is increased up to 6. Secondly, for the sake of computational burden alleviation, a candidate vector selection algorithm is introduced in such a way that the number of candidate vectors is reduced to only 7–9. Then, the problem of neutral-point voltage fluctuation which is inherent to the three-level NPC inverter is investigated, and a reasonable replacement method incorporating the redundancy characteristic of small vectors is developed. Finally, experimental results verify that the system robustness against mismatched parameters can be greatly enhanced by use of the proposed control scheme; meanwhile, steady-state performance and dynamic-state response of the system can be improved to some extent as well.