Basically, if you are doing any work that requires heavy numerical processing, Cython is the way to go. On the other hand, I was playing with it to do some basic text processing and the improvements were negligible.
Even for numerical processing, it depends on where/how you write your loops. numpy just offers bindings to optimised C libraries and loops using map(..,..) should be much faster than using for.
As far as text processing is concerned, it seems like the python code is just a nice interface to the underlying compiled library and hence there isn't much difference.
http://www.perrygeo.net/wordpress/?p=116
Basically, if you are doing any work that requires heavy numerical processing, Cython is the way to go. On the other hand, I was playing with it to do some basic text processing and the improvements were negligible.