Task modelling has enabled the building of models of human activity for a long time. In the early years, pencil and paper where the only means available to build task models from task model notations. Because of the lack of computed constraints, task models often did not conform to the notation. To solve this problem, some tools were designed by authors in order to help users create, modify and save correct models that conform to the notation syntactic rules. However, understanding the full semantics of task models appeared difficult for practitioners. The dynamic aspects of task models could only be understood "in the user's head". New tools, named simulators, emerged to solve this problem. They allow to "run" or to "simulate" task models and to record scenarios. This execution fulfils the semantics of task model operators, which define the task dynamic semantics. Simulators can be used in many ways such as understanding model semantics, verifying or validating models, building valid scenarios, etc. In this article, we describe and compare currently available and maintained task model simulators, and explain the different usages of these tools, according to user goals and qualifications. Then, we explore the different challenges for these tools to exploit the complete semantics of task models.