You are right, #2321 made the numbers a bit tighter:

With a small string:
./python -m timeit -s "s=open('INTBENCH', 'r').read()" "s.split()"
-> Unpatched py3k: 23.1 usec per loop
-> Freelist patch: 21.3 usec per loop
-> PyVarObject patch: 20.5 usec per loop

With a medium-sized string:
./python -m timeit -s "s=open('LICENSE', 'r').read()" "s.split()"
-> Unpatched py3k: 406 usec per loop
-> Freelist patch: 353 usec per loop
-> PyVarObject patch: 314 usec per loop

With a long string:
./python -m timeit -s "s=open('Misc/HISTORY', 'r').read()" "s.split()"
-> Unpatched py3k: 22.7 msec per loop
-> Freelist patch: 24 msec per loop
-> PyVarObject patch: 20.6 msec per loop

stringbench3k:
-> Unpatched py3k: 266 seconds
-> Freelist patch: 264 seconds
-> PyVarObject patch: 249 seconds

Regarding your benchmarking suggestion, this would certainly be an
interesting thing to do, but I fear it is also much more than I'm
willing to do...

I'm going to post the updated patches.