Seems more like margin of error differences, except for CS:GO and maybe a few very select others

This is the best summary I could come up with:

Released late on Friday was the much anticipated SteamOS 3.5 preview for the Steam Deck with ongoing work around HDR and enhancing color management, VRR for external USB-C displays, various platform issues resolved, auto-mounting external storage, and more.

With SteamOS 3.5 it also means some lower-level OS upgrades too like moving to the Linux 6.1 LTS kernel.

For those wondering about the performance impact of going from SteamOS 3.4 stable to the SteamOS 3.5 preview release, here are some early benchmarks on the Steam Deck.

In being curious about the performance of the Arch Linux powered SteamOS 3.5 update, I ran some initial benchmarks over the weekend looking at the impact of SteamOS 3.5 preview as of 15 September compared to SteamOS 3.4 stable.

SteamOS 3.5 also moves from KDE Plasma 5.26 to 5.27 on the desktop along with having various other package upgrades for its Arch Linux base.

A variety of games were benchmarked on SteamOS 3.4 stable vs. 3.5 preview plus some additional Linux CPU workloads in just getting an overall idea for the performance of this forthcoming SteamOS upgrade for the Steam Deck.

The original article contains 264 words, the summary contains 186 words. Saved 30%. I'm a bot and I'm open source!

@ylai how?

Did they read their own article? because I did and it seems to be "within margin or error" or "measurably a tiny bit worse", apart from CSGO (which suggests a testing error if anything).

TLDR Clickbait.