AI APIs
Is OpenAI down?
Live OpenAI status, measured independently. We probe 4 OpenAI endpoints from monitoring regions on every continent and record every outage we detect.
OpenAI is up
Our probes are reaching OpenAI normally from every region we check from.
Checked every minute from 6 regions on every continent.
- Uptime, 24 hoursUptime, last 24 hours
- Uptime, 7 daysUptime, last 7 days
- Uptime, 30 daysUptime, last 30 days
- API99.98% uptime, 30 days6 regions, 81 to 320 msUnknownp50median response200 msp95slowest 5% of checks340 ms500 ms250024 h ago12 h agonow
- Developer platform100% uptime, 30 days6 regions, 20 to 79 msUnknownp50median response44 msp95slowest 5% of checks131 ms100 ms50024 h ago12 h agonow
- API documentation100% uptime, 30 days6 regions, 264 to 1042 msUnknownp50median response506 msp95slowest 5% of checks839 ms1000 ms500024 h ago12 h agonow
- DNS100% uptime, 30 days6 regions, 2 to 29 msUnknownp50median response10 msp95slowest 5% of checks34 ms20 ms10024 h ago12 h agonow
Daily strip:99.9% or better95–99.9%below 95%no data yet
Get notified when OpenAI goes down
Follow these 4 checks from your own workspace and get alerted through Slack, email, webhooks, or on-call the moment OpenAI goes down. They don't count against your monitor quota. Free plan, no card required.
About OpenAI outages
The OpenAI API powers chat, summarisation, embeddings, and agent features in a vast number of products. OpenAI incidents are frequent enough that any product depending on it needs a fallback, and they often present as elevated error rates or latency rather than a hard outage.
We probe the API, the developer platform, and its documentation. Consumer surfaces are behind aggressive bot protection, so we do not probe them; the official page covers them.
What we check
Every row is a real monitor in upti.my, running from each of our regions. A failure in some regions marks the row degraded; failures from every region mark it down and open an outage.
| Check | Target | Expects | Interval |
|---|---|---|---|
API The OpenAI API is not answering with its normal authentication response. Completions and embeddings fail. | httpapi.openai.com/v1/models | HTTP 401 | 1 min |
Developer platform platform.openai.com is unreachable. API keys and usage dashboards are affected. | httpplatform.openai.com | Any HTTP response | 2 min |
API documentation platform.openai.com/docs is unreachable. | httpplatform.openai.com/docs | Any HTTP response | 5 min |
DNS The API hostname is not resolving. | dnsapi.openai.com | A record resolves | 1 min |
For component-level detail and incident write-ups from the vendor, see OpenAI's official status page.
Recent OpenAI outages we detected
An outage is recorded when a probe fails from every monitoring region and closed when it recovers. Partial and single-region problems show as degraded above but are not listed here.
OpenAI is up. Is your OpenAI integration?
The incidents that actually hurt are usually not global outages. They are the parts of your own system that depend on OpenAI and fail quietly. Here is what teams that rely on OpenAI monitor on their side.
Latency and error rate, not just reachability
OpenAI degradations usually mean slow or failing completions while the endpoint answers. Probe a real completion from your own backend and alert on latency thresholds.
HTTP monitoring with latency alertsFallback routing
When the primary provider degrades, a workflow can switch to a fallback model automatically and open an incident.
Workflow monitoring and automationOpenAI status: frequently asked questions
A 429 is rate limiting on your account or organisation, not an outage. Widespread 500s and timeouts during an incident are declared on the official page, usually as elevated error rates on a specific model. Our probe confirms the API endpoint is reachable and authenticating.
Look at the per-region results above. If our probes fail from every region, OpenAI has a real outage and it will appear in the outage list. If they succeed everywhere, the problem is most likely on your side: your network, DNS resolver, a browser extension, a corporate proxy, or your own integration. If only some regions fail, OpenAI may have a regional incident that does not affect you.
Every probe runs at least every five minutes from each monitoring region, and most run every minute. This page refreshes its data every minute. Uptime figures are calculated from all checks across all regions.
Official status is updated by people after an incident is confirmed and often lags the first failures by many minutes; our probes react within a minute or two. In the other direction, we only see public endpoints, so an internal incident may not show on our probes. Use both.
Yes. Use "Notify me about OpenAI" above: your upti.my workspace follows these exact checks, free plan included, and every outage opens an incident there and routes alerts to Slack, email, webhooks, or on-call. Nothing is duplicated and it doesn't use your monitor quota. Better still, also monitor your own dependency on OpenAI: the API route, webhook receiver, or scheduled job that calls it, so you learn about the failure that actually affects your users.
Related
Other ai apis
upti.my is not affiliated with OpenAI. OpenAI and related marks are the property of their respective owners and are used here only to identify the service being measured. Results reflect the reachability of public endpoints from our monitoring regions and may differ from the vendor's own reporting.