Many induction motor models are available in the literature for modeling, simulation, control as well as fault diagnosis. In this paper, novel induction motor models are formulated based on stationary, rotor and synchronously rotating reference frames. All these proposed models are compared and analyzed in terms of their diagnostic relevance to stator inter-turn faults. Ability to generate distinct residual signatures is a key for model-based diagnostics. Performance of these models in term of their ability to diagnose a fault is compared and best suited models are suggested based on a discriminatory ability index proposed in this paper. As continuous-time extended Kalman filter (CTEKF) is the most commonly used estimator for state estimation, the same has been considered. Computer simulations are carried out for a 4-hp squirrel-cage induction motor using MATLAB. The results prove that all induction motor models do not perform equally well when it comes to fault diagnosis. © 2016 IEEE.