We believe that with a proper dynamic JIT (like pypy) a version of clojure running on a dynamic VM can outperform its JVM and CLR counterparts. Aside from that, there are many Python libraries like PySide (Qt GUI), numpy, scipy, and stackless that do not have JVM counterparts, or at least the Python implementations are easier to use and learn. clojure-py will integrate tightly with thy Python VM and will be able to use all of these libraries.
So now you're probably asking yourself, "Why should I care?" Well, Clojure has strengths in areas that Python doesn't, so a combination of the two could bring the best of both worlds.
Metaprogramming: Clojure's Lisp heritage makes it extremely suitable for macro-based metaprogramming because it follows the "code is data" philosophy.
Persistent Immutable Data Structures: Clojure has a set of immutable lists, vectors, sets and maps, and since they can't be changed, a new collection is created each time something is added or removed.
Clojure-py is available for download on Github. There's also a roadmap for the project that has a list of items marked "isolated change," which are changes designed to be easy for newcomers, if you'd like to help the project along.