Prevent automatic GPT downgrade in chat GPT when using voice

When using voice chat for a while, the chat GPT model at some point gets downgraded from 4.0 to 3.5.

When using voice chat, can we have a prompt or allow us to have an option to prevent model downgrading? For example if we have reached a limit, let us have the option to stop communication so that we can return at a later point.

Alternatively, a work around would be to give us The option to upgrade The chat back to 4.0 if it has been downgraded. If we have reached a limit, then we can upgrade it later.

Certain conversations are important to keep in 4.0 and I would at least like this option to prevent downgrading.

It has caused me to have to start over the chat and go through a very long context over again.

Thank you!

1 Like

I am also frustrated by this issue, and 100% agree we need a solve.

If it helps, here is a prompt I just created that helped me port a conversation into a new ChatGPT 4.0 while preserving the instructions and the latest output.

Provide a detailed, clear, step-by-step summary of the process we’ve discussed, along with any important context, revisions and iterations along the way. This detailed summary will be used as a guideline in subsequent interactions with ChatGPT models. Your output should be worded as a detailed prompt or brief to provide to future ChatGPTs, so make sure it reads as instructions. As the last part of your instructions, provide the latest output of the task (and make it clear that this is a starting point for future discussions).

Thanks for the suggestion. Unfortunately, for one, The chat that this happened to that was downgraded, gives me an error each time I query it in 3.5.

Also, asking for a summary doesn’t really seem sufficient for the case that I use it for. Over time, the long running chat has begun to behave differently. It feels there is nuance to the context where tokens form that is hard to capture with a text summary. Or that might not be accepted in a new chat without more gradual building of a complex context.

I feel that a better solution is if we could copy an existing context to a new 4.0 chat. Either that or upgrade the chat back to 4.0.