untold sapphire
Expected result:
My goal is to continue using GPT-4 once again on a ChatGPT message thread which was originally GPT-4, but has been switched to GPT-3.5, not to convert an originally GPT-3.5 thread to GPT-4.
Actual result:
I have a long and detailed chat history with GPT-4 I've been using for app dev; after switching from GPT-4 to GPT3.5 following a token limit, I cannot switch back.
My setup:
ChatGPT Plus on OpenAI's website
Safari Browser, 2021 MacBook Pro, macOS Ventura 13.2.1
Steps to reproduce:
- Open ChatGPT, become a ChatGPT Plus member, and open a chat with GPT-4.
- Use GPT-4 until the token limit has been reached (currently, this is just 25 messages in 3 hours).
- Upon reaching the limit for GPT-4 tokens, click the button that gives you the option to continue using GPT default model, (GPT-3.5 by default).
- A disclaimer message pops up (with no buttons to go back to GPT-4) which says: The previous model used in this conversation is unavailable. We've switched you to the latest default model.
- Continue chatting, noting that now the icon has changed from Black (indicating GPT-4) to Green (indicating GPT-3 or 3.5). The returned responses are notable worse, have less ability to refer to context, and are not cohesive or useful.
Things I've done to try to resolve this problem:
- Wait until the token limit has been refreshed/removed after the specified wait time, which didn't revert the model downgrade
- Refreshed browser tabs/Signed out and back in/Tried using different browsers/Tried using different devices etc to no avail
- I tried scrolling back up to a previous prompt which was responded by GPT-4 to edit my original prompt using the prompt edit button, only to note that the new response to my edited prompt forced the model to switch to GPT-3.5 and not continue using GPT-4
**I've seen other people posting the same problem here: **https://discord.com/channels/974519864045756446/1085762065274249226
slate skiff
mossy radish