Abstract: Due to the inherent characteristics of behavior prediction and direct manipulation of converter switching states, the finite control set model predictive flux control (FCS-MPFC) and its variants can ensure fast dynamic responses for control variables. However, high switching frequency is normally required and significant degradation in steady-state performances can be observed when the average switching frequency is decreased. In order to solve this problem, this paper proposed a two-step variable action period model predictive flux control (TS-VAP-MPFC) algorithm for permanent magnet synchronous machines. With the proposed control scheme adopted, the vector action duration of selected voltage vectors could be flexibly regulated among two-step prediction horizon, which introduced an additional control freedom. The experimental results demonstrate that, at the same average switching frequency, improved steady-state performances and excellent dynamic responses can both be expected with the proposed algorithm adopted.