Hi Preeti,
You can't get a single model working best in all of the cases as the model learning totally depends on the data, which is feed, as you mention that the datasets are different, which will result in different learning of the same model as the weights learned will be different as per the data fed. As a good practice, we will generally create separate models for different kinds of data, but if you have such an use case where you need to use a single model, then you need to make some sort of tradeoff.
Hope it helps!
