As you can note over previously posted results, much depends on opening book. To be fair, we must use every book, but repeating each opening for both colors. And if instead you want to test your book, both sides must use the same engine, ore results are pretty random, since all SF derivative are all on the same strenght range smile  But Proteus is now learning, so I expect even better results against Kayra 1.3.

Good Proteus SF 007 JBE performance against SF-PB 010422! 43gRbwnY

TCEC 21 Jeroem Noomen select super final PGN book, each position repeated for both sided.
Tablebases Syzyzy 6 Man. TC 3 mins + 2 secs / all on Mac mini M1 8 CPUs calculating 6.5 Mn/s.