Likewise here in UK with our own in house developed tool chain using gpt-realtime and it has been unresponsive since at least the last 5 hours. Have attempted to reach Justin Uberti on X but no response - quite early in US currently.
Hey everyone, Apologies for the inconvenience this has caused but we have our engineering team actively reviewing this issue and we should have an update very soon. Huge apologies!
Hey @OpenAI_Support
Can you communicate on your status page please ? So that we receive updates.
Thanks you
+1
400 on SIP direct to openai, webhooks not hit.
SIP isnt really stable. last issue was in november. i would recommend to install a fallback to websockets. at leat i did. quality sucks but at least you have no downtime
FYI, OpenAI has escalated this issue internally.
How are we to trust the status page when a total outage since yesterday is not mentioned?
Hi @curt.kennedy
Is there any estimated time to solve this problem?
We’re talking about a 17h service down…
It is set to High Priority and they have reproduced it internally, I don’t see any mention of a timeline on this one.
Thanks @curt.kennedy
so, due the fact that there’s no mention in status, could you send here any feedback?
thanks
Hey everyone, We have our status page updated and everyone can follow the updates here: https://status.openai.com/incidents/y29b1rax. Apologies again!
“Elevated error rates on Realtime API SIP endpoints”
Cut the crap!
The service has been completely down for hours.
I love how they call it degraded performance when it is 100% down. Elevated error rate? Technically true from 0 to 100% lol
I didn’t write the title. But at least it is officially recognized on the status page now.
Thanks @curt.kennedy @OpenAI_Support
Seems sorted now
Thank you
Started working for us a few minutes ago as well. Hopefully OpenAI improves reliability over SIP (and response times in case of failures), because otherwise it’s very hard to build production products on this setup, and we’d need to look for alternatives. SIP had issues a couple of months ago for several days, and today we experienced a total outage for 20 hours. Thanks ![]()
It is still a load of crap. It was an outage. OAI has seeked to marginalise its customers by making them think it wasn’t an outage and your stupid bot told me it was our problem. Clearly it wasn’t just us having the problem. There wasn’t one successful call during this period.
If you want to retain your customers, your web site should tell the truth.

