Yet to find a model that cross-model review doesn’t find a bunch of things wrong with. I’m running simultaneous review with whichever of Grok4.6/GLM5.3/Fable/Sol didn’t write it, and each model tends to find items the others didn’t.