Response.image_generation_call.completed not coming through?

API Bugs
,
1

Unless I’m doing something wrong - I send off an image_gen request through API which works great (or at least, used to work great) - and I get the partials back like this:

Screenshot 2025-10-02 at 7.06.58 PM
Screenshot 2025-10-02 at 7.06.58 PM2078×310 70.3 KB

But for some reason, we no longer get the image_generation_call.completed event? Can anyone else confirm this issue or is it our implementation?

1 Like
2

I’m noticing the same behavior; streaming does not seem to be sending the completed event.

3

Thanks for raising this. I am also seeing the same behavior. We receive the interim partial events, but the image_generation_call.completed event never arrives. It looks like a recent change on the API side rather than an implementation issue.

4
  1. We used to rely on the completed event as well, and it seems to have stopped appearing. Our code has not changed, so this might be a current bug in the event stream. Hopefully OpenAI can confirm.
5

We still have to do some testing, but I'll post here once fixed! ETA early next week.

Thanks everyone for your patience.

2 Likes
6

So cool - thank you.

While you’re at it - if you don’t mind - can you check that initial event surrounding the image_generation. Can’t remember exactly which one, but I do know that it was something like 15-20 seconds to get that initial indication that yes, the model is generating an image.

In other words, the ChatGPT “starting up” seems to happen WAY sooner than API…not sure if thats intentional or not, but if there’s a way to send it sooner…:folded_hands:

Thanks again for your attention and work.

1 Like
7

This has been resolved as of Nov 6 - sorry for the late update and thanks again for the reports!

1 Like
8

I’m still encountering this bug (model: gpt-4o-mini).
I’m receiving two partial images but never a response.image_generation_call.completed event.
In the response.output object, the image status is always “generating”.

Can you confirm the fix is deployed ?

EDIT: I did upgrade the openai nodejs SDK to 6.9.1, model to gpt-5-nano, same bug

9

I can confirm that I’m still only seeing “generating”

10

I haven't been able to repro anymore, but perhaps my settings aren't correct! Would you mind sharing your repro steps? e.g. model, tools, prompt, streaming status, etc.


Thank you!

1 Like