Nuitka is a Python compiler.
It's fully compatible with Python 2.6, 2.7, 3.2 and 3.3.
You feed it your Python app, it does a lot of clever things, and spits out an executable or extension module.
Free license (Apache).
Right now Nuitka is a good replacement for the Python interpreter and compiles every construct that CPython 2.6, 2.7, 3.2 and 3.3 offer. It translates the Python into a C++ program that then uses "libpython" to execute in the same way as CPython does, in a very compatible way.
It is somewhat faster than CPython already, but currently it doesn't make all the optimizations possible, but a 258% factor on pystone is a good start (number is from version 0.3.11).
In the future, Nuitka will be able to use type inference and guessing by doing whole program analysis and then applying the results to perform as many calculations as possible. It will do this - where possible - without accessing "libpython" but in C++ with its native data types.
It will also be possible to integrate "ctypes" based bindings without the normal speed penalty (the compiled program will call the C++ library in C++ directly).
And finally you will be able to use a "hints" module to tell Nuitka about type information that it cannot guess.
It's a road with milestones.
Well milestone 1, feature parity has been reached for Python 2.6, 2.7, 3.2, and 3.3 mastered. There is no way, this could be any better.
With milestone 2, it is considered reached. Nuitka is more than 2 times faster than CPython because of this. While this is certainly not going to give the anywhere near the biggest gains possible, these are solid improvements that can only help the goal to be as fast as possible. Even in final Nuitka, not all objects will always be correctly type inferenced and then the quality of milestone 2 will be able to still be faster than CPython.
Aiming at milestone 3, the constant folding and propagation is in place. Some control flow optimization are applied too. But this only at a start. It will give the big gains when types are inferenced.
For milestone 4, first steps have been undertaken to achieve "type inference". The results are encouraging, but will need a lot more work, before they can be made the default approach.
And milestone 5+6 are not even started. So way to go until we hit the "future".
In the meantime you will find its latest version here.