This study presents potential and limitations of utilizing the current functionality of selected multimodal large language models with reasoning capabilities to solve structurally complex mathematics-based problems that involve graphical outcomes. As a part development of study materials and tests for introductory university math courses, we focus on exploring capabilities of large language models to process and generate graphical input and output in three-dimensional space. Incorporating the geometric perspective into the problem solution based on calculations supports and improves deepening students' understanding of mathematical concepts of solved tasks. As for comparison of model´s generated results to the graphical representation provided by specialized software, we discuss both the benefits and issues of this approach. We also evaluate the acceptability of generated outcomes in terms of fulfilling the requirements given by user´s query. Additionally, we highlight the importance of well-formulated prompts in ensuring high-quality results.

Keywords: Large language model, mathematics, three-dimensional visualization, artificial intelligence.