Interpreted Matlab
As an interpreted language Matlab has several benefits over compiled languages in regard to mathematical operations. By far the most prominent of these is ease of debugging. With interpreation there is no need for a separate debugger or clunky in-code diagnostics.
Furthermore, the ability to test code snippets using interactive one-liners is a great aid to rapid development without the need for extensive (and often expensive) IDEs.
The price to be paid for all this flexibility is, of course, speed. While much of the number crunching in Matlab is performed in heavily-optimised routines, the glue is necessarily going to be much slower and while a skilled programmer can mitigate these effects they cannot be nullified completely. The trick is to use MEX files for those functions where speed is critical or where execution takes up most of the run time (as evidenced by profilers.)