Two modeling techniques of microwave devices are proposed to generate neuro-fuzzy-based models. These techniques use the adaptive neuro-fuzzy inference system approach, which compensates the error between an initial coarse model and an electromagnetic simulator (or measurement data). The aim of these techniques is to generate accurate models from a set of fuzzy if-then rules and coarse models. Models so obtained could be integrated in a toolbox of any commercially available computer-aided design tools for radio frequency/microwave circuits. Results with artificial neural network and neuro-fuzzy based models are listed and discussed for a microwave tunable phase shifter.