Whilst ChatGPT broad AI models into the zeitgeist, the long-term landscape for AI will be a varied set of models trained to b competent at tasks for particular domains.
Choosing between these models may be done automatically based on the desired output, however for some circumstances providing users a selection provides flexibility for accuracy or experimentation.
Another advanced user pattern, to fully utilise a model Selection feature requires some understanding of what models exist, their intended purpose & what data they have been trained on.
It's likely most user interface are designed with a particular goal defined, so the appropriate model would be pre selected. However, as the list of available models grows this option may become more relevant.
Consider if your users would benefit from having this option. Could they not only experiment with their prompts for better outputs but with various models as well.