Is GPT-4o through ChatGPT secretly using GPT-4o-mini for simple prompts? The "sycophantic behavior" fix might be masking a deeper shift

GPT-4o in ChatGPT does not switch between the full GPT-4o model and the lighter GPT-4o-mini model depending on the complexity of the prompt.

But sometimes, they coud be a bug or some testing by OpenAI developers.

When you feel it happens, go to this link:

GPT 4o works only on web search mode - #5 by polepole

And try what is explaining there.

I think you don’t need a theory about it.

There is no a secret or a hidden switching mechanism.