except for the R(10^8) case which is now taken by the sequence implementation indicating a low constant
Another aspect is that my timings were taken from inside the python process, using the timeit module, in order to time only the computation itself and avoid including the startup time of the CPython interpreter, which is considerable. If measurements are taken from outside the python process, this interpreter startup time will be included.
boy is the Python3 interpreter fast!
You're a funny one.
You might want to run a benchmark that actually stresses bigints.
I think the higher-order versions will serve that purpose.