my guess is they are and always have been using TINA to collect capture samples and data , that does a large part of the heavy lifting , in the case of certain amps that maybe the captures don't full handle , then they are likely sending in a dev to get further measurements and comparing to the amp and do som post EQ tone matching , but the core models are probably very similar to something like MIKKO in which there are so many captures, that they put them all together so you have a smooth transition in the range of the controls .
That very different to Fractal component level modelling and getting into things like impedance modelling behaviours