Kant writes in his Groundwork of the Metaphysics of Morals:
The above three ways of representing the principle of morality are
fundamentally only so many formulae of the selfsame law, one of
which of itself unites the other two within it. However, there is yet a
dissimilarity among them, which is indeed subjectively rather than objectively practical, namely to bring an idea of reason closer to intuition
(according to a certain analogy) and thereby to feeling. (4:436)
The three mentioned ways are the Formula of the Law of Nature (4:421), The Formula of Humanity (4:429) and the Formula of the autonomous will in a Kingdom of Ends (4:434), the last one containing the other two (see Timmermann's Kant’s Groundwork of the Metaphysics of Morals: A Commentary, Cambridge UP, 2007, pp. 105-6, 109-11).
All of these three are "closer to intuition (according to a certain analogy)", which is paraphrasing that while a will governed only by morality is "an idea of reason" (which has no correspondence in experience, strictly speaking), i.e. we do not have a concept of it, we can use analogies like "law of nature", "end in itself" and "autonomy in a kingdom of ends" to get an idea of what this could actually mean.
This, Kant states, does not imply an "objectively practical" difference, but only a "subjectively" one. In short: They are objectively the same, they just have a different "feeling" to them.
But in the end, there is only one categorical imperative, which is the reason why he is able to state:
But
in moral judging it is better always to proceed by the strict method,
and make the foundation the universal formula of the categorical imperative: act according to the maxim that can make itself at the same time a universal law. If,l however, one wants at the same time to obtain access
for the moral law, it is very useful to lead one and the same action
through the said three concepts and thereby, as far as can be done,
bring it closer to intuition. (4:436-7, bolded PK)
TL;DR
Kant's answer is that if it helps you, you may use all versions as they are objectively the same. But strictly speaking, there is only one single categorical imperative and applying this is always the "best" (most rational) way to judge morality.