I think it is important to understand the modeling process Naughty Dog uses to craft their game. Naughty Dog always begins with a hi resolution sculpt built in Zbrush. I think on ps4, I heard these are million polygon plus models. They extract the normal map from this model. Later this information is applied to the low res model to transfer some of the details of the hi res model to the low res model. This hi res model is scaled down for the in game cut scenes.
Uncharted 3 used around 30,000 -50000 models for both cutscenes and realtime models. Naughty Dog uses the same model for the cut scenes and for the real time model so the transition is more seamless.
However, the lighting, texture resolution, and shadowing are dulled down for the in game model. What this means that while the real time model looks great when viewed from a certain distance away from the camera, the model loses detail as you get closer. They also LODs for entities other than Drake. These means as the camera is further away from the entity, it can be constructed with less detail to save resources.
While the modeling and assets look to be on the same level, it is pretty obvious the lighting and texture resolution is dialed down from the e3 teaser. It's great that most people won't notice as Naughty Dog has always hoped for this even with their previous games.
The original teaser's lighting and shadowing rivaled what you could achieve using Maya with Vray or MentalRay. Still in comparison to other games, it looks great especially with Naughty Dog's amazing facial animation.
Here is how each model looks like in each stage using Laz and Drake models for the previous Uncharted games.