We can do even better some of the time - for example, we can get to ~few nanosecond timing precision for pulsars like J1909-3744 [1], which is equivalent to measuring (changes in) the path length to that pulsar to a few meters or better.

[1] https://arxiv.org/abs/1406.4716