@DgYWFT I completely agree but these things have schedules for these reasons.
Guys, problem is living now?
Hello @VeitB, are there any news regarding this issue? Based on your experience how long does it take for resolving similar issues for the OpenAI team?
@VeitB have you received any updates from the team? Thanks
I’ve asked the team for an update and will share any news as soon as I receive it.
For now, the most constructive step is to migrate to the Responses API sooner rather than later, as Assistants will be sunset in a few months.
Community feedback on the migration has been very positive, so I think this is the best solution.
Hey everyone, Apologies for the delay. Can someone please open a support request with us and provide the HAR file in the support case. Please make sure to include your forum username and link to the forum discussion in the support request so we can track it. Furthermore, please share the case id or case subject here with us so we can use the information supplied. We will escalate this to our engineering team promptly and get this resolved as soon as possible but since Assistant API is deprecated [More info] I will request if you can use response API and start migration with your own pace, while we get this fixed. We are really sorry about this issue.
This is a complete platform-wide outage of service that has been ongoing for two weeks. It does not need someone to offer you free consulting or to figure out what a HAR file is and for you to waste their time making people do extra work. It is immediately actionable.
Assistants - Replication steps
Purpose built vision Assistant, existing,
- with prompting right there about images that now OpenAI took a clue from and even copies into their own system message model injection.
- Assistant configuration
- Platform site dashboard image input
- “upload image” is broken and does nothing, so we resort to “link”
- Constructed message with image for vision
- API failure, same as has been persisting
This needs no association with any user ID or tracking, or trying to take the issue off the forum so that you can again restart all over with bot responses that do nothing.
Or you can simply let us know plainly (from all appearances that are presented), that while you have the ability to send with “OpenAI Support” on the forum, you are completely powerless to take initiative and communicate with “real” OpenAI … at all.
Perhaps only in comparison to Assistants, taken alone…
Getting a “server_error” issue when using the Assistant API.
Using an assistant with the gpt-4.1 model.
Created a thread with an image. After running the thread, the thread status started at queued and then changed to failed with the error text “server_error”.
I’ve been facing this issue since 23 Feb 2026.
If your are using Flex Tier, all started failing totally for me
UPDATE - We ended up upgrading to the new API. It looks like OpenAI does not have development resources allocated to fix bugs for a product that will be deprecated within a few months.
Hey Everyone, Apologies. Our team is looking to fix this issue. We are actively looking into this and should have an update very soon. Really sorry for the experience you are facing.
Any update when this will be fixed? This has been a bug for almost a month now!
Been like a month, they wont fix it, your best to just convert your code and or migrate to another service (we migrated, and are porting to another provider, this was a clear sign of how they will handle future changes).
Hey everyone, Apologies for the inconvenience. We have deployed a fix for this issue and this should now be resolved. Incase, you are still facing the issue please open a support request (support@openai.com) and our team will take a look into individual setup.