Well, if you want to nitpick about that, then you have to define what an LLM is.
Because: yes it is an LLM property to be non deterministic.
However it is not an "artificial neuronal network" property, as those are deterministic ... albeit the determinism can not be described. Unless you want to go hardcore nitpicking: this (insert a number) X billion parameters describe exactly the ANN (which is true, it does), and hence describes it determinism.
Modern AI are not LLMs, never where actually. They are based on LLMs. The whole thing is much more complex then an LLM or an ANN.
I suggest to start reading up on it ... or be left behind.
You are a computer scientist, you should be fascinated by them, instead of disregarding them as play tools.