This article proposes an ultralocal predictive surface-mounted permanent magnet synchronous motor (SPMSM) model-based predefined time sliding mode predictive speed control (UL-SMPC) to achieve exceptional disturbance rejection and tracking performance in SPMSM systems. First, an ultralocal predictive SPMSM model is given, incorporating a time-varying disturbance term and an adaptive control gain. Second, the control gain and disturbance term of the SPMSM model are decoupled and identified, respectively. A control gain optimizer is devised to estimate the control gain, and a predefined time reaching law-based generalized proportional integral observer (PT-GPIO) is developed to estimate the disturbance for each control period. The parameter tuning principles for the control gain optimizer and PT-GPIO are analyzed. Third, a cost index is defined using the predefined time reaching law-based sliding mode surface. Finally, UL-SMPC is synthesized by minimizing the cost index. Experimental results verify the outstanding robustness and tracking performances of the proposed method.