Model within a model?

How can I take two feature vectors as outputs from two inputs, store them, and compare a feature vector from a third input to the first feature vectors within one model? End result is dataset C is 80% similar to dataset A 20% to dataset B, etc.