Models are predictions. I drop an apple; it falls to the ground. I do that often enough and I can build a model that something draws objects to the centre of the Earth. I notice that, if I generalize the model, it can also predict the Moon going around the Earth. Turns out I cn do that rpetty well with my model. As long as that prediction is reliable, I have a pretty good model. When we looked at the bending of light around massive objects, we see our basic model is no longer entirely predictive. We modify our model to one of bending spacetime. Our model is now far more predictive. It still has problems. At no time does not model declare what is "really" happening. And at no time does it ever declare anything is proven.