The 4o-mini model is failing for assistants. 
Anyone else having this issue? I had to switch the model on all my clients’ bots because they all stopped working. I’m going with 4.1 mini for now—let me know if that’s the best option.
2 Likes
You are correct. gpt-4o-mini is failing on Assistants. Kicking off an Assistants thread returns nothing but a closed connection.
gpt-4o succeeds, as well as gpt-4.1-mini.
gpt-4o-mini works on Chat Completions and Responses.
I have the same error Run has been ended with an invalid status “failed”. Sorry, something went wrong.
Origin
OpenAI (ChatGPT, Sora, DALL-E, Whisper)
I’m facing the same issues.
same issue here, suddenly stopped working today
Its happening for me for the last 8 hrs. If i change the model to 4o its working 
Anyone following this thread/issue I contacted OpenAI support and provided a link to this thread. Hopefully their engineers will respond soon.
Aqui também parou tive que modificar para usar gpt-4.1-mini para voltar a funcionar, mas estamos notando diferenças nas respostas e na qualidade da tratativa.
1 Like
I personally wasn’t able to reproduce the issue but send a heads-up to the team from here as well.
Hoping this gets resolved for you soon!
It looks like the gpt-4o-mini model is working on Assistants again.
But then, it only takes the first “working” output to exhibit a previously-reported issue in the platform site - “null” being shown for text data that was output by the AI model.
For simply what should be a bullet point in markdown:
I could probably give a run down at least half-a-dozen ongoing API issues right now, but. why bother…the topics just scroll off the screen without any attention being paid. The API is just getting ridiculous and falling apart, probably from those vibe-coded commits.
1 Like