The time complexity of n - 1 and n & m is not O(1) in Python. Moreover, builtin functions are much faster in Python. That's why the solution from the link is significantly faster than yours on relatively large inputs.

Try to run your solution with n = (1 << 500000) - 1 and this solution with the same n.

For relatively small numbers there's no difference. If you test it with huge inputs like (1 << 65536) - 1, it's super slow compared to others.

that's right for Python 3 (this solution passes the kata in Python 2)