Hi ​@Ranit Efron, welcome to the Community! 🎉

Sorry to hear you’re running into those timeout errors. I did some checking in the OpenAI community and it seems that GPT-5 is intentionally slower than the GPT-4o model which is likely why it’s timing out: https://community.openai.com/t/gpt-5-is-very-slow-compared-to-4-1-responses-api/1337859

That said, I spotted that we’ve got a bug report open for these timeout errors that are affecting certain models, like the GPT-5 one. I’ve added you to the list of folks being affected by that issue. I can’t make any promises around when you can expect it to be fixed but you’ll be sure to get an email notification from us as soon as it is. 

For the time being, I’d recommend using the GPT-4o model, instead of GPT-5.

Let me know if there’s anything else I can help with in the meantime. 🙂

I am constantly hitting this error when using AI by zapier steps. I have large LLM prompt due to the context being included and I cannot break them into smaller steps. Please fix this or give us an option to extend the timeout for certain steps. 

Sorry to hear you’re also running into these timeout errors ​@Zac. I just checked and the team has been making some progress on this but the bug report is still open at the moment. I’ve added you to the list of users being affected by this to ensure you’ll get a notification once it’s been marked as resolved. Thanks for your patience in the meantime. 🙏

Hey, I am experiencing the same issue. Is there any update?

The bug is still unresolved at the moment ​@danstenson. I can’t give any sort of update as to when you can expect it to be fixed but I’ve added you to the list of folks being impacted so you’ll be notified as soon as it’s sorted. 

However, I’ve just seen a note that enabling Autoreplay allowed the errored Zap runs to go through successfully a little while later. So it might be worth trying to replay any previously errored Zap runs to see if they’re now able to complete. If anyone here gives that a try, let me know how it goes!

Hi, I can confirm on my end and vouch that this is a large and existent issue at this time.

Zapier devs, please kindly resolve this!

I’m so sorry you’re also being impacted by this and for the disruption it’s causing, ​@CryptAlchemy. I’ve added you to the list so you’ll be sure to get an email notification once it’s been sorted. 

Would definitely suggest trying to replay any errored Zap runs in case that helps in the meantime.

In case you’re only seeing the timeout issues in testing and haven’t published your Zaps, I recently learned the below from someone at Zapier support. In my case, I just ran a test in GPT 5-nano so I could connect the test output to subsequent steps, and then switched it to GPT 5 and re-published without re-running a test, and it worked okay in live runs.

Test runs have shorter timeout windows (typically around 30 seconds) to provide quick feedback during setup. This is why you're seeing the 504 CloudFront errors when testing with GPT-5 mini—the model needs more processing time than the test environment allows.

Live runs have much longer timeout limits (up to 10+ minutes for AI steps), which is why your full replay succeeded even though it took 5-10 minutes to complete. This extended timeout is designed to accommodate the longer processing times that more sophisticated AI models require.