Design patterns are important for improving code reusability, maintainability, and overall software quality by providing standardized solutions. Most part of design pattern detection methods, such as static analysis and graph-based techniques, face challenges due to variability in pattern implementations, scalability issues, and ambiguity in design rationale. In addition, extracting and processing full code bases is complex and not always feasible due to resource limitations. Integrating design pattern detection with inference through Large Language Models (LLMs) and MDE techniques may overcome some of these limits by producing abstractions of large code bases. In this paper, we propose to align with Model-Driven Engineering (MDE) principles, helping in the automation of extraction of code bases in the form of UML models and injecting them into an LLM-based design-pattern recognition flow. We use the P-MART repository for design pattern detection to evaluate the solution’s effectiveness. We compare different LLMs using models with and without comments. According to our findings, LLMs are able to identify a variety of Gang of Four (GoF) design patterns using UML models as input, but with limits, particularly when combining multiple patterns.