This article deals with the principles of organization of hybrid architecture of cluster computing based on the use of computing opportunities of the CPU and GPU with support for GPGPU. The advantages and disadvantages of this method of calculation have been analyzed, and the feasibility of realization of parallel programs based on the use of a hybrid architecture of cluster computing has been demonstrated. A software framework that can be used to develop the applications has been implemented on the basis of the proposed programming model. In order to demonstrate the opportunities of practical use of the hybrid architecture and to analyze its potency based on the proposed software framework, implemented program of block Fourier Transform of high-resolution images. The results of processing, substantiation and comparison of the results are given in the last part of the article.