I re-ran them :-) https://github.com/BurntSushi/ripgrep/tree/master/benchsuite...

TL;DR ripgrep has gotten faster in important areas since the initial set of benchmarks (the proper comparison there would be https://github.com/BurntSushi/ripgrep/tree/master/benchsuite...). The key reasons why are because it grew a parallel directory traverser, and its line counting got vectorized courtesy of the bytecount[1] crate. ucg has gotten a little faster in some cases, but the general conclusion of "ripgrep is the fastest" is still correct.

[1] - https://github.com/llogiq/bytecount