Hi,
I’m a paying Pro subscriber and for the past two days, I haven’t been able to use inference or test any models without getting Internal Server Errors. Every generation attempt fails mid-process. I’ve tried different models, cleared cache, different browsers, same result every time.
Your status page says everything is operational, which clearly isn’t the case.
Looking through this forum, I can see I’m far from the only Pro subscriber experiencing this. There are dozens of threads from paying users who suddenly can’t use what they’re paying for , models stop working overnight, quota gets eaten by failed generations, and the documentation on what Pro actually includes is intentionally vague.
At this point it’s hard not to see a pattern. People subscribe to Pro expecting a working service, and instead they get the same errors as free users with no clear communication or resolution.
I need this fixed or at the very least a clear explanation of what’s going on. I’m based in the EU, and if this continues, I’ll be reaching out to consumer protection authorities, as I’m not receiving the service I’m paying for.
Thanks.
1 Like
Was that Space working before the last two days?
If we know which Space it is, we might be able to tell whether the error is caused by HF or by the creator’s code.
If you say “the last two days,” there’s a strong suspicion that an internal server error (the kind that doesn’t show up on the status page) has occurred.
e.g. https://huggingface.co/proxy/discuss.huggingface.co/t/my-hugging-face-space-keep-starting/174616 https://huggingface.co/proxy/discuss.huggingface.co/t/unban-me-please/174586
However, there are also many cases where the Space’s code simply stops working… (which, given that we use Python, is a constraint that’s bound to happen to some extent).
Unless we know which one it is, no one will be able to fix it.
Hi there,
Really sorry to hear you’ve been dealing with this it’s understandably frustrating, especially as a Pro subscriber expecting reliable access.
A few things to help clarify and (hopefully) move this forward:
1. Status page vs real-world issues
The status page usually reflects system-wide outages, but it doesn’t always capture regional issues, account-specific problems, or partial service disruptions. So it’s possible for things to appear “operational” while some users (or regions like parts of the EU) are still experiencing errors.
2. Internal Server Errors during generation
What you’re describing—failures mid-generation across multiple models—can sometimes be caused by:
-
Temporary backend instability or load issues
-
Region-specific routing problems
-
Account/session-related glitches (especially after plan changes or renewals)
Since you’ve already tried different browsers and cleared cache, it likely isn’t a local issue on your end.
3. Quota being used on failed requests
You’re right to flag this—failed generations shouldn’t meaningfully impact your usage. If they are, that’s something support definitely needs to review and potentially credit back.
4. What to do next (important)
At this point, the best step is to contact official support directly with details so they can investigate your account:
Include:
-
Approximate timestamps of failed requests
-
Models you were using
-
Screenshots or error messages (if possible)
-
Confirmation that it’s happening across browsers/devices
This helps them trace logs and identify whether it’s a backend or account-specific issue.
5. On Pro plan expectations
I understand your concern here. While Pro plans generally provide priority access and higher limits, they don’t guarantee zero downtime. That said, repeated failures like you’re describing especially over multiple days are not typical and should be addressed.
1 Like