Hello Sims
This is a very good question that doesn’t have a clear cut answer.
In the hierarchical model, the core is performing high speed routing and switching to get things from place to place, the distribution layer is applying policies and filtering, and the access layer is providing you with port density and user connectivity. Based on this, and based on most textbooks and CCNP study material, QoS is said to be most commonly applied at the distribution layer.
However, Layer 2 switches at the access layer can perform both marking and queuing and in general it is good design to apply QoS as close to the equipment or services for which you need to guarantee network resources as possible. In a live network that may very well be in the access layer on your switches, or even at the phone port.
I hope this has been helpful!
Laz