Technically random can have probabilities associated with them.. Casual speech, random means equal probabilities, or we don’t know the probabilities. But for LLM token output, it does estimate the probabilities.

Greedy decoding isn't random.