zumatools.

Popular right now

Add Prefix / Suffix to LinesPaste one item per line, then add text around each line, join them with a separator,…Tally CounterTap the big card to count up, use the minus button to correct mistakes, and add as…Love CalculatorEnter two names and get a playful compatibility score — the same pair of names always…Photo to Pencil SketchConvert any picture into a hand-drawn pencil sketch directly in your browser, with…Tic Tac ToePlay against a friend on the same device or challenge the computer on easy or…Random Pair GeneratorPaste names below and split them into random pairs or groups of three, shuffled fairly…
Free tools. Right here in your browser.esc to close

Hugging Face · Service status

Is Hugging Face down right now?

Hugging Face is the main hub for sharing and downloading machine-learning models, datasets and demo apps (Spaces). It also sells inference through Inference Providers and Inference Endpoints.

Live status · Hugging Face

Checking live status…

Reading the official status page, our reachability checks and recent user reports.

Official status page
—status.huggingface.co
Our reachability check
—Checked every minute from our server
User reports
—From visitors to this page

Having trouble with Hugging Face?

Tell us what’s wrong. One report per person every 30 minutes; nothing else is collected.

Hugging Face in the last 24 hours

Loading the last 24 hours…

Reachability (our checks)No data
  • Reachable
  • Partly reachable
  • Not reachable
  • No data
Response timeNo data
User reportsNo data

If Hugging Face isn’t working for you

When the live checks above look normal, the problem is usually on your side or limited to one feature. These are the fixes that help most often with Hugging Face:

  1. Check the right component

    The status page tracks the Hub, Git hosting and serving, Inference Endpoints, the Spaces proxy, CDNs and Jobs separately. Model downloads, a Space and the website can each break on their own, so check the component behind the part you use.

  2. A sleeping or paused Space is not an outage

    Free Spaces go to sleep after a period without use and restart when someone visits, which takes a moment. The owner can also pause a Space. A Space that shows a build or runtime error is a problem with that app, not with Hugging Face.

  3. 429 Too Many Requests

    The Hub rate-limits API calls, file downloads and page views over short windows, and unauthenticated traffic gets the lowest limits. The most common fix is to set an HF_TOKEN so every library and tool that downloads from the Hub sends your token.

  4. Inference Providers: credits and partner providers

    Inference Providers routes your request through Hugging Face to partner providers. Free accounts get small monthly credits. If one provider is having trouble, name a different provider or use a selection policy such as :fastest — a single provider failing isn’t a Hugging Face outage.

  5. ZeroGPU quota

    ZeroGPU Spaces have a daily GPU-time quota that depends on whether you’re signed in and on your plan. Signing in or waiting for the reset fixes quota errors; it isn’t a platform outage.

Official Hugging Face sources

Questions about Hugging Face outages

Is Hugging Face down or just one Space?

Usually just the Space. Spaces are apps run by their authors and can be asleep, paused, rebuilding or crashed while the Hub works normally. This page combines the official status page, our own reachability checks every minute and user reports.

Why do model downloads fail with 429?

You’ve hit the Hub’s rate limit, which is lowest for unauthenticated requests. Set HF_TOKEN so every library sends your token; paid plans get higher limits.

Why did my Inference Providers requests stop working?

Check whether your monthly credits are used up, and whether the provider serving that model is having problems — Inference Providers routes to third-party companies. You can switch providers by adding a provider name or selection policy to the model ID.

What’s the difference between Inference Providers and Inference Endpoints?

Inference Providers is serverless, pay-per-use access to models run by partner providers. Inference Endpoints are dedicated deployments you manage yourself, and the status page tracks them as their own components.

How this page decides

The answer at the top combines three signals: what Hugging Face reports on its official status page, our own check that Hugging Face is reachable (every minute, with response time), and reports from people on this page. Problems are flagged when the official source reports an incident, our checks start failing, or reports rise well above the usual level for Hugging Face. It’s an independent page — not affiliated with Hugging Face. See all AI services.

Check another AI service