In software testing stage, faults can be revealed and then diagnosed to identify defects that caused it. Tests should ideally be applied from the unit level to the higher level of software, such as system testing. In one of these levels resides GUI (Graphical User Interface) testing. Ensuring the correct operation of the GUI on the state of its elements after various user events is as important as the other layers tests, since GUI is a direct way to interact with the application, being the feature that most influences how the experience will be qualified by the end user. This paper proposes a Model-Based Testing (MBT) approach using high-level Petri Nets (RP) to represent graphical user interface. Petri Net is a modeling tool and a mathematical specification language that graphically defines the structure of systems, specially the concurrent ones. An important feature of RPs is that they can be simulated, allowing one to observe the behavior of the system and to obtain the generation of test cases from the paths executed in the simulation. The generation of test suites for GUI from the Petri Net model was investigated. For this, we considered the relationship between user actions and resulting states in the GUI, realizing how an RP can model them. A support tool was developed so that, from the simulations of Petri Net, test suites were generated in C plus plus language, making it possible to auto-run them on a study software. Finally, the Mutation Analysis test criterion, which measures the effectiveness of the suite generated from RP, was employed as a means of validation of this work.