32/64 bit float is the native float type on the cpu you are running on.. hence you can't really claim "native" support for larger types.. if you mean native in terms of being able to write a+b for a numeric type, then that eliminates all languages that don't have operator overloading..