> In future are we going to have same model for everything?

A model that knows more in general, will often be better at specific tasks. e.g. If you ask a model to "make a program that estimates the annual production of a solar installation", it needs to have been trained on a lot more than just Python code.

You might combine a general world model with a python coding model in that case. Not sure if it's better, just saying.