Proxyway had zenrows at 70.39% at 2 req/s dropping to 31.76 at 10 req/s, and they noted concurrency caps probably affected those results

Also the bright data part got mixed up. That 93.14% is zyte's score from 2025, bright data's 97.90 is from 2024 with different targets. Just source label each number and clarify which benchmark year it came from so the benchmark reference doesn't oversell the comparison