Is Hugging Face down?
Right now
No. Hugging Face is operational.
No incidents reported.
Read from status.huggingface.co · every 300s
For agents
proceed
Proceed.
GET upbutler.com/status/huggingface/status.agent.json
MCP: upstream_status {"service":"huggingface"}
The Hub (model and dataset downloads), Inference Endpoints and Spaces. A Hub outage breaks cold starts that pull weights at boot.
What Hugging Face reports
Operational
Taken as is from Hugging Face's own status page API. Incident titles and component names below are theirs.
Open status.huggingface.coUpButler's independent probe
Hub API is answeringHTTP 401 · 132 ms
One request without credentials to huggingface.co every two minutes. An answer (the expected "not authenticated") shows the edge, TLS and routing work. It does not prove that authenticated calls succeed.
Checked · up since 2026-10-10 01:23 UTC
Hugging Face components
8 operational
Current status by service
- Huggingface HubOperational
- Git Hosting and ServingOperational
Inference Endpoints
- Inference Endpoints UIOperational
- Inference Endpoints APIOperational
Spaces
- Spaces ProxyOperational
CDN
- AWS CDNOperational
- Google cloud CDNOperational
Jobs
- JobsOperational
Hugging Face incidents, last 90 days
No finished incidents on record yet since 2026-10-10. UpButler keeps what Hugging Face publishes for 90 days from the day it starts tracking a service.
Tell your outage from Hugging Face's
Declare that your project depends on Hugging Face. When one of your monitors fails while Hugging Face has an incident that began around the same time, UpButler tags your incident "likely upstream" with their incident linked, says so in the alert and the AI report, and tells your on-call agent not to change code.
npx upbutler init finds the dependency from your env var names and packages. The catalog, dependencies and upstream-aware incidents are on every plan, including Free.
# upbutler.yaml
dependencies:
- huggingfacecurl -s https://upbutler.com/status/huggingface/status.agent.json{
"schema": "upbutler.upstream-status/v1",
"service": "huggingface",
"verdict": "proceed",
"status": "operational",
"reason": "No incidents reported",
"retryAfterSec": null,
"incident": null,
"probe": {
"state": "up",
"statusCode": 401
}
}Questions
- Is Hugging Face down right now?
- No. As of 2026-10-10 05:11 UTC, Hugging Face is operational. This page reads the status again every 5 minutes.
- Where does this status come from?
- From Hugging Face's own status page API at https://status.huggingface.co. UpButler reads it once for everyone (not once per visitor) every 300 seconds with conditional requests, and shows it unchanged: incident titles and component names are Hugging Face's words.
- What does UpButler's own Hugging Face check prove?
- Every two minutes UpButler sends one request without credentials to the Hugging Face Hub API (huggingface.co). Any answer from the edge, usually HTTP 401, shows that DNS, TLS and routing work. It does not prove that authenticated requests succeed, so it can say "reachable" while Hugging Face reports degraded service. It is useful the other way round: when the probe fails before their status page changes.
- How can my agent or app check Hugging Face status?
- Fetch https://upbutler.com/status/huggingface/status.agent.json: a small JSON document with a verdict (proceed, retry, fallback or pause) and retryAfterSec. No key is needed. The same answer is an MCP tool, upstream_status {"service":"huggingface"}, on https://upbutler.com/mcp/public.
- How do I know whether an outage is mine or Hugging Face's?
- Add "huggingface" to the dependencies of your UpButler workspace. When one of your monitors fails while Hugging Face has an incident that began around the same time, your incident is tagged "likely upstream" with their incident linked, and your on-call agent is told not to change code.
More in ai models and tooling
UpButler is not affiliated with Hugging Face. Status and incident text belong to Hugging Face and are read from their public status source; the probe result is UpButler's own measurement.