Post
104
Does anyone think a **Pro sub is worth it just to duplicate basic CPU demo Spaces** to avoid queue? Especially when ZeroGPU is a thing.
Through 2026 I built mostly CPU-only:
Luminia **~45 CPU Spaces across 3 orgs**, each *running* 3-5 at once, But now.. sorry to everyone following those orgs, currently I can't create new Spaces anymore.
In July 2026 HF quietly changed who can create free CPU Spaces:
- only **Pro** ($9/mo) or **Team** ($20/user/mo, org) can create new Gradio/Docker Spaces
- free users **can't duplicate** CPU Spaces anymore
So everyone shares the same instance and queue. Pretty rough for long-running jobs.
> "You already have CPU Spaces, why not consolidate them to free up slots?"
Some Spaces load the model at startup and keep it in RAM instead of per run. More models in one Space makes every run slower, and not everything can be merged in a meaningful way.
> "Pause some Spaces, paused ones don't consume a slot."
Pause would be fine if users could restart it themselves, but that only works while a Space is *sleeping*, which needs 48h inactivity, and mine are used daily. No manual sleep trigger. Transfers to free orgs are blocked entirely.
> "Just buy Pro?"
That only covers my own account, with limited slots. Each org has its **own CPU quota**, so to keep building in ONE org I'd need Team.
> "Use Static HTML Spaces."
Free, but not every device can run AI in-browser, and support is limited.
---
I tried HF email and "Send feedback". No answer.
As a creator I don't think this should exist at all. It goes against the open-source ethos and kills creativity, visibility and accessibility. **No reason given, cost or otherwise**, and staff aren't engaging with 6+ complaint threads on discuss.huggingface.co.
Please @huggingface revert this policy.
Through 2026 I built mostly CPU-only:
In July 2026 HF quietly changed who can create free CPU Spaces:
- only **Pro** ($9/mo) or **Team** ($20/user/mo, org) can create new Gradio/Docker Spaces
- free users **can't duplicate** CPU Spaces anymore
So everyone shares the same instance and queue. Pretty rough for long-running jobs.
> "You already have CPU Spaces, why not consolidate them to free up slots?"
Some Spaces load the model at startup and keep it in RAM instead of per run. More models in one Space makes every run slower, and not everything can be merged in a meaningful way.
> "Pause some Spaces, paused ones don't consume a slot."
Pause would be fine if users could restart it themselves, but that only works while a Space is *sleeping*, which needs 48h inactivity, and mine are used daily. No manual sleep trigger. Transfers to free orgs are blocked entirely.
> "Just buy Pro?"
That only covers my own account, with limited slots. Each org has its **own CPU quota**, so to keep building in ONE org I'd need Team.
> "Use Static HTML Spaces."
Free, but not every device can run AI in-browser, and support is limited.
---
I tried HF email and "Send feedback". No answer.
As a creator I don't think this should exist at all. It goes against the open-source ethos and kills creativity, visibility and accessibility. **No reason given, cost or otherwise**, and staff aren't engaging with 6+ complaint threads on discuss.huggingface.co.
Please @huggingface revert this policy.