The ChatGPT desktop app currently defaults new chats to the last model opened.
For example:
- Start a new chat and set it to o1-preview, engage in a conversation.
- Switch the model to 4o in that same chat.
- If you open a new chat, it defaults back to o1-preview.
This is problematic for several reasons:
- Users may unknowingly continue using o1-preview, quickly depleting their weekly allowance.
- This behavior potentially increases strain on OpenAI’s systems, as more users unintentionally generate queries with o1-preview instead of using a default like 4o.
To address this, the default behavior for new chats should always revert to 4o, ensuring users avoid accidental usage of limited models and reducing unnecessary server load.
New chats should never default to a model that has a limited allowance.