Why would you only update the original base model and not simply continue fine-tuning? (This is a rhetorical question of course, and there are multiple empiric

Why would you only update the original base model and not simply continue fine-tuning? (This is a rhetorical question of course, and there are multiple empirical ways to confirm the majority of your answer is "hallucinated," not least of which is the fact that OpenAI keeps rolling out new versions and variants with different capabilities.)