Abstract
The article discusses the possibilities of using multigenerative neural network models to solve applied problems in design. The aim of the study was to compare the results of design project generation performed by ChatGPT, Gemini, and Copilot neural networks based on specified text scripts and initial visual data. As part of the experiment, two types of tasks were formed: the creation of a new small architectural form and the refinement of an existing design project for a computer audience. The evaluation of the results was carried out by an expert group of designers according to a system of criteria, including artistic, functional, ergonomic and economic parameters. The data obtained showed that multigenerative neural networks are capable of generating competitive conceptual solutions that vary in the degree of technological complexity, design expressiveness, and feasibility. The most balanced results were demonstrated by solutions focused on a combination of visual expressiveness and practical applicability. The conclusion is made about the prospects of using multigenerative approaches as a tool to support project activities in design.