The Hyper benchmark uses multiple threads. When I ran the benchmark, I modified the version found in Hyper's github to use both the old multi-threaded scheduler and the new multi-threaded scheduler.

I probably should make that clear in the article.