Helios is the first (to our knowledge) modeling language for global optimization using in- terval analysis. Helios makes it possible to state global optimization problems almost as in sci- entific papers and textbooks and is guaranteed to find all isolated solutions in constraint-solving problems and all global optima in optimization problems. Helios statements are compiled to Newton, a constraint logic programming language using constraint satisfaction and interval anal- ysis techniques and their efficiency is comparable to direct programming in Newton. This paper presents the design of Helios, describes its theoretical foundation and semantic properties, sketches its implementation, reports some experimental results, and compares Helios to other modeling languages and direct programming in Newton.