The problem of solving complex and large-scale tasks using computers is discussed, where modeling is the main method of solving tasks. The advantages and disadvantages of classical programming considered as the main tools of computer modeling are analyzed. A model-Theoretic approach to solving the problem of computer-based solution of complex tasks, called semantic modeling, is described. The advantages of the proposed concept are analyzed from the perspective of the ability to solve complex large-scale tasks by constructing hybrid models combining various approaches and modeling methods using oracles, which are basic predicates of the original model. This feature allows you to integrate models of various nature, for example, digital and analog, continuous and discrete, logical-mathematical and probabilistic, mathematical and heuristic, etc. based on semantic modeling. The problem is to create a single semantic protocol that allows integration of models of different nature.