>> I've been mostly using GLM-5.3 on low reasoning, with great results.

Bro come on.