In traditional and e-learning the tests play an important role to track the learners’ progress. In this regard, the paper deals with the generation of test with different levels of complexity directly associated with particular evaluation. An algorithm for the generation of tests with different levels of complexity using multi-criteria optimization is proposed. It is important to note that all test items could be considered as an element of a level in an educational game or gamified environment, and their weight influences the complexity of the game. The results show that the proposed model allows the generation of tests not only with different levels of complexity, but it also provides additional flexibility in respect of the selected number of questions and their weight. The proposed algorithm could be realized as a separate module for etesting, or it could be integrated in the learning management system.