Unfortunately, the answer is "it depends". When it makes sense I prefer to dimension the model the same way it will appear on the drawing. Often there are reasons to dimension the two differently. Usually these reasons boil down to attempts at reducing drawing clutter.
The most common example of different model/drawing dimension schemes I use are putting drawing dimensions either as ordinate off a single datum and/or in a hole table; while the in-model dimensions are often hole to hole, or patterns (which are essentially a type of hole-to-hole dimension). My design intent is expressed via GD&T feature control frames, not by how my dimensions are located. Most of my dimensions don't have tolerance info, due to placing that in feature control frames for most features.