Does anyone know if there is a dedicated model which makes Claude output nore human readable and less slop?

Lately it became load-bearingly-reality-difficult to not only read, but to comprehend the Claude output