I am using 4o’s API to analyze some images. It used to work fine, but today it replies with:
Run failed:
Sorry, something went wrong.
Does this happen to anyone else?
Same problem here. It worked well until today. When sending images it returns the error you mention. I hope it gets fixed soon!
It is model-specific and when using an image file: gpt-4o fails, however, gpt-4-turbo is successful.
gpt-4o with vision works on chat completions.
Another weekly reminder to get your app off of Assistants…
For an existing assistant, I added a file to an auto-created vector store in the playground like you. The question was answered, no error.
Then with an image, upload reported success, and the file_id is included as thread message and message call is repeated back. But then the run failed.
Inspecting the calls, the run progress is polled, transitioning from in_progress, to:
last_error Object { code: "server_error", message: "Sorry, something went wrong." }
(Inspecting the calls, we also see the playground defeats the whole idea of assistants preloaded with instructions: sending what’s in the instruction text box as new run instructions)
Same problem here. Entire day getting ERROR - Error details: {
“code”: “server_error”,
“message”: “Sorry, something went wrong.”
} while using gpt-4o with images via API.
1 Like
It appears that there is an issue with gpt-4o-2024-08-06.
Image analysis was successful with gpt-4o-2024-11-20 and gpt-4o-2024-05-13.
The following is an excerpt from the runs response.
"last_error": {
"code": "server_error",
"message": "Sorry, something went wrong."
},
"model": "gpt-4o-2024-08-06",
"last_error": null,
"model": "gpt-4o-2024-11-20",
"last_error": null,
"model": "gpt-4o-2024-05-13",
1 Like
Yes. I have the same issue. Unfortunately for now OpenAI don’t show this problem on https://status.openai.com/ so I could assume that they don’t start to fix it yet.
1 Like
We are using gpt-4o with the assistant API and are receiving this errors very often since yesterday. Worked quite well before
Hey all,
We’ve been using OpenAI assistants with the code interpreter to generate charts for a while now. We just started running into some unexpected errors today.
Note the following thread:
We’re seeing this “something went wrong” error consistently when using an assistant with the gpt-4o model.
As I mentioned, we haven’t changed this configuration in quite a long time, and we just started running into issues today. The error result from our runs just looks like this:
"last_error": {
"code": "server_error",
"message": "Sorry, something went wrong."
},
which doesn’t give us much to go on.
Was wondering if anyone is running into a similar issue, or if someone could provide some insight. Interestingly, using gpt-4o-mini seems to solve the problem.
Yes! I have been having the same problem, it fails instantly if I try to use vision or code_interpreter
I reached out to OpenAI directly and they’ve acknowledged an issue on their end and have engineers looking into it. Will follow up if I hear any more updates.
I am also have mixed issues with OpenAI requests hanging on o3-mini and 4o-mini using the NodeJS openAI API.
Insanely frustrating as I spent hours debugging since their outage page shows no issues.
In me experience, you can use “gpt-4o-2024-11-20” version and it seems to fix the problem for now
Sorry for the troubles here. This is now fixed and Assistants should be working properly again, as of 17:45 UTC.
2 Likes
We are still getting
{ code: 'server_error', message: 'Sorry, something went wrong.' }
Maybe you want to take a look at thread thread_2IGPZ4EqQB7UBrsWJftcFQdO
Thanks!
Even when it doesn’t run into an error it randomly ignores structured output instructions and answers in plain text.
Switching to model gpt-4o-2024-11-20 seems to fix both issues.