What Badiou does in The Concept of Model is give an intra-mathematical example of how mathematics, as pure production of differential marks, is able to gain traction across different domains (the example is model theory, where a mapping is established between a domain of statements and a domain of sets such that the latter "models" the former), so that a physical science can become axiomatised or a philosophical "meta-ontological" discourse can "place itself under condition of" innovations in set theory (obviously the "innovations" discussed in Being and Event are all pretty old hat, but part of Badiou's argument is that philosophers still haven't really caught up with them).
So I would say that the statement "math is only descriptive" seems radically false to me. It's not even that. Or, it is only "descriptive" to the extent that other domains attempt to formalise their own knowledge by mathematical means. If they didn't - if they knew nothing, or were content to rest in "wild empiricism" - then mathematics would not in any sense describe their contents. Physicists may "apply" mathematics within their domain, but this neither makes physics "applied mathematics" nor mathematics "only descriptive" of the physical universe (not least because it is capable of being used in the description of inexistent universes, universes that have yet to come to be).