Hi everyone,
I’ve been trying to run my space using ZeroGPU recently, but I keep encountering issues where it fails to secure an allocation/reservation (frequently getting timeouts or no GPU available).
Is anyone else currently experiencing the same issue, or is there an ongoing outage/high load on ZeroGPU? If there are any known workarounds or best practices to handle this, I’d appreciate the help.
Thanks!
I’ve been suffering from it as well. Couldn’t figure out why and it has lasted for like 6 hours already.
Same here! I also noticed that only a few top Spaces are currently running compared to usual, so it definitely looks like an issue on their end.
Cannot do anything other than just waiting for them to fix 
Same here , to whom contact for that ?
Yes, exactly. I’ve had random luck with 1/20 or so attempts. Just wasting away my daily quota.
Looks like Spaces are starting to regain the spinning fan with number indicator and they work fine.
Not having the same experience as you with that one. Have been running into “No GPU was available after 60s.” all day.
Have ran through nearly 15 minutes of my daily quota without managing to connect to a gpu even once.
What is going on?
Yep, even when the space shows the spinning fan indicator with a number, I still can’t connect
Same here, PRO account, all day.
Two data points that might help narrow this down:
It is NOT high load. The ZeroGPU panel on the Space itself shows
“This Space is currently using 0 RTX Pro 6000 Blackwell”, CPU 0%,
RAM 5/104 GB — completely idle. The “Top ZeroGPU Spaces” leaderboard
shows only single-digit concurrent usage (7, 6, 4, 4, 2). The fleet
looks empty, yet nothing gets allocated. This is not saturation —
the scheduler simply isn’t handing out GPUs.
It is NOT account-, quota- or priority-related. I ran the exact same
request with NO token at all (fully anonymous, no account association)
and got the identical “No GPU was available after 60s”.
My quota is not exhausted (22.1 / 40 minutes used), yet even Kokoro-82M
(a tiny TTS model that normally takes ~1 second) cannot get an allocation.
Tested on 5 different Spaces — Kokoro-TTS-Zero, LTX-2.5,
MiniMax-H3-Turbo-Lora, Wan_2.2_I2V_14B-Clean, wan555 — all identical.
Also confirming what others said above: failed attempts DO consume daily
quota. I burned 22 minutes today with almost nothing generated.
This is an EPIC FAIL and a good reason to leave and cancel PRO subscription. Addition: FAILS to create NSFW contents. Another EPIC FAIL.
Yes, I’ve been seeing the issue since yesterday evening. Like other users, I’ve tried multiple models and have been unsuccessful in getting a ZeroGPU to connect. Frustrating to say the least. Just tried another attempt and it timed out as well…
It worked fine on Aug 20th.
Same here. I give them one more chance before cancelling subscription. This is not what I want to get for my money. Additionally, Krea 2 fails to generate nsfw content, so another reason 
They urgently should disable the zero GPU count for failed attempts. Who got THAT stupid idea anyway? Especially since Huggingface is far from stable at all. Makes no sense and is very bad advertising in general.
I’m sure it wasn’t a conscious decision on their part, the whole thing is an unexpected outage for which the usual pipeline just didn’t account for
I’ve always thought the same thing. ZeroGPU quota shouldn’t be deducted if it doesn’t actually connect to the GPU.
Same here.
When I change the ZeroGPU runtime to CPU Basic and then switch back to ZeroGPU, it sometimes fixes the issue…
I can’t quite tell if it’s just because the system is overloaded or if it’s some kind of glitch.
@hysts
I sent multiple emails no response yet Any other way to contact ?
Does someone already get the solutions or explanation whatever is happening right now?
same here, till now.