Domain-Specific Languages (DSLs) play a vital role in software development, enabling the concise expression of domain-specific concepts and requirements. In this study, we propose a novel approach leveraging Large Language Models (LLMs) to assist the DSLs modelling starting from natural language description. Our solution is a proof of concept where Model Driven Engineering (MDE) is revisited taking advantage from the power of generative AI. Starting from human friendly description and domain modelling language document type, LLM-based system extracts relevant domain knowledge and builds the corresponding DSL model. Such a model is then validated through an iterative process. We applied our proposal to several case studies from different application domains including software engineering, healthcare, and finance. Furthermore, we consider a wide range of existing LLMs usually adapted for code generation. We also study the effectiveness of our solution through multi-criteria experimental evaluation. Lastly, the results demonstrate the feasibility and efficiency of our LLM driven MDE for DSL development, and then advancing domain-specific modelling practices. By doing so, we would enable the developer to save time and effort for further tasks such as functional properties’ verification. A demo as well as a web application for our developed solution are available online via the following link https://alaouimdaghriahmed.github.io/demo-ecore-gen/ .