Model-driven engineering (MDE) provides abstraction and analytical rigour, but industrial adoption in many domains has been limited by the cost of developing and maintaining models. Large language models (LLMs) help shift this cost balance by enabling the direct generation of models from natural-language (NL) descriptions. For domain-specific languages (DSLs), however, there is the risk that LLM-generated models may be less accurate than LLM-generated code in mainstream languages such as Python, due to the latter’s dominance in LLM training corpora. We investigate this issue in the domain of mathematical optimization, focusing on AMPL – a DSL with established industrial use. We introduce EXEOS, an LLM-based approach that derives AMPL models and Python code from NL problem descriptions and iteratively refines them using solver feedback. Using a public optimization dataset and real-world supply-chain cases from our industry partner, Kinaxis, we evaluate how generated AMPL models compare with Python code in terms of executability and correctness. An ablation study across two LLM families shows that AMPL is competitive with – and sometimes better than – Python, and that our design choices in EXEOS improve the quality of generated specifications.

Keywords: Model-Driven Engineering (MDE), Large Language Models (LLMs), Automated Model Extraction, Domain-Specific Languages (DSLs), Mathematical Optimization.