This exact thing is happening to me, including the AI erroneously saying it is still GPT4 and then acknowledging it’s really 3.5 when pointed out. This behavior of downgrading is new as of yesterday for me on the Android app.
More frustrating is that all help channels are useless. The FAQ have nothing on the topic, no biggie, but their directions for how to log in to get help don’t work - nowhere to send a message. I tried a support email and it doesn’t go through anymore.
Bottom line - OpenAI has downgraded the service and has no support available for Plus subscribers.
Glad to hear someone else is having the same issue!
An interesting thing I found out yesterday was voice is using the standard model so if you start a chat on your phone with voice you can use it without it breaking
So far nothing has changed with this issue and I just installed and reinstalled the app with no change in the model or behavior. The new model is still being used and it crashes after one or two uses. Any other ideas from the community? Can someone from OpenAI respond?
I have seen staff, other than Kilpatrick, respond twice this year to posts. I believe Kilpatrick is off until January, and he also rarely responds, so the chances are low.
From a phone, open the ChatGPT app and send a message to GPT4. After receiving a message response, you can click the drop-down at the top of the screen and select See Details. The model in use will show.
@grandell1234 , I opened a ticket 12/20 with screenshots. I replied back 12/21 asking for an update. I received a response back from Edel@openai 12/24 with zero information about what I emailed about.
——— Edel @ OpenAI’s response: ———
Hello,
Thank you for reaching out to OpenAI support.
We apologize that you are encountering the error. Our servers may experience intermittent errors during periods of high traffic, which may affect free users. We recommend trying again in a few minutes or starting a new chat.
You may also try the following:
Try clearing your browser’s cache and cookies
Disable VPN (If you are using one)
Use an incognito browser or a different internet browser/computer to see if the issue still persists, as a security add-in or extension can occasionally cause this type of error.
Disable cookie blockers
Try connecting to another network
Please inform us if you have experienced any other errors.
This will help our engineering team troubleshoot the issue.
Best,
Edel
OpenAI
——- End Email Response———-
Again, this was their response to a detailed email I provided with screenshots of the issue.:
———- My email to open a I Suport ———
Gpt4 crashes after one or two requests. It says it is using model
gpt-4-retrieval-msearch-ctx
I have not found anything on the internet referencing this model. Can you help me understand what it is and why it is using a different model? Did I sign up for a beta program? If so, how do I report bugs? It crashes after a few requests down to 3.5 with no warning. I actually have a chat where it pretended it was still in gpt4 mode even though it showed 3.5. I appreciate any clarification.
I look forward to hearing back from open-air support soon! Thank you!
——- End of my email to open AI ———-
I was happy to get any response back though. Thanks Edel! I replied back with
Hi Edel,
I don’t believe that is the issue and you have not addressed the
gpt-4-retrieval-msearch-ctx
model. What is it? Why do I have it in use? Am I on a beta model? If so I was not informed. As a paying customer I would love more info. Thank you Edel!
———
Still waiting on a response. Will keep this thread updated as I learn more
I just received another reply from Edel. I am not sure what it has to do with the issue at hand.
Hello Zane,
Thank you for reaching out to OpenAI support.
Thanks for reaching out! We wanted to let you know that we recently made some changes to our site that you might appreciate.
Now, every generation you create will automatically be saved by default in the “All generations” tab in “My collection.” You can access this tab on both web and mobile here.
This change is retroactive, so you’ll be able to find any past generations you may have forgotten to save. It’s a great way to see how your DALL·E style has changed over time!
If you have any questions or concerns about these new changes, please don’t hesitate to reach out to us.
While I can’t tell exactly what version of the model the web page ChatGPT is using it is also stating that it is unable to use the model used initially and the default model is in use. I assume that is the same model crashing on web as well. HELP!
A little further analysis appears to show that when the model crashes it stays on GPT4 still which is better than mobile which crashes to 3.5. Still interesting but at least usable.
Note the small text "The previous model used in this conversation is unavailable. We’ve switched you to the latest default model
this same email thread with support has numerous attachments including everything captured here. I even included a link to this post. This is frustrating.
Hello Zane,
Thank you for reaching out to OpenAI support.
Would you kindly provide us with a screenshot of the page that is not functioning properly on your phone? This will assist us in determining the underlying issue.
I believe a human response has been received from OpenAI via the chatbox.
" Hello,
Thank you for reaching out to OpenAI support.
We want to extend our sincere thanks for bringing the recent issue to our attention. Your feedback is invaluable to us, and in response, we immediately escalated your concern to our engineering team. We can confirm that they are actively working on a fix.
At this time, we are unable to provide a specific timeline for when this issue will be resolved. However, we assure you that we are committed to resolving it as swiftly as possible. Our team understands the importance of this matter and is dedicating their resources to finding a solution.
We will make sure to keep our community informed and updated as soon as a fix is implemented. In the meantime, your patience and understanding are greatly appreciated as we continue to work diligently on this issue.
If you have any further questions or need additional assistance during this period, please feel free to reach out to us.
Best,
Genesis
OpenAI Support"
Fingers crossed! I volunteered to join the beta program to try to help the cause (and hopefully get a better communication vector for feedback to them). Fingers crossed on that as well.
A quick update on the gpt-4-retrieval-msearch-ctx model! It is still broken as hell. Here is a video of me trying to continue using the mobile app on a chat started on their website.
The crazy thing is that Edel continues to claim that they are aware of the issue
“ Hello Zane,
Thank you for reaching out to OpenAI support.
We found that your concern has already been raised to our internal team. We cannot guarantee the time frame when it will be addressed. The only thing we guarantee is that this has already been reported, and they work hand in hand to enhance our service and address all customer feedback.
Thank you so much for sharing your thoughts on our products! We value your feedback greatly and we have made sure to take note of your comments. We will be sure to review them thoroughly within our team.
If you have further comments or suggestions, please do not hesitate to contact us.
Best,
Edel
OpenAI Support”
But when I ask why they still report status as operational.
“ Has this been resolved? I am still showing status as reported as operational but I am still unable to use a working version of GPT4. Please update the status and let me know what the ETA is to have this issue resolved.”
Of course they do not respond to this targeted question.
Continued frustration for this service. I am owed a refund for the last month.
. I hope our feedback helps