In this work, the impact of iron loss on speed estimation performance of the conventional rotor flux error-based model reference adaptive system (MRAS) estimator is investigated in case of induction machines (IMs). In addition, two improved MRAS-type estimators are proposed to reduce the estimation error caused by iron loss. The first approach takes into account iron loss resistance by its nominal constant value. But in the second case, iron loss is frequency dependent. All three estimators are compared by simulations. The results show that the MRAS estimators with iron loss compensation can reduce the speed estimation error and the frequency dependent iron loss compensation method provides the highest accuracy.