Is OpenAI down?
Operationalchecked 1 minute ago
We reached OpenAI from all 10 regions we check.
As of Oct 5, 2026, 8:16 PM UTC, OpenAI answered RealUptime's probes in all 10 regions that returned a reading.
Last outage: 11 min on Aug 28, 2026 (API, from Europe, US-East and US-West).
OpenAI reports: Minor incident (“Partial System Degradation”)
OpenAI reports an active incident: Elevated error rates in Codex Cloud (all systems operational, since Oct 5, 2026, 5:58 PM UTC)
OpenAI's own component breakdown, as they publish it
- Chat CompletionsOperational
- ResponsesOperational
- AgentsOperational
- DecisionsOperational
- Fine-tuningOperational
- EmbeddingsOperational
- ImagesOperational
- BatchOperational
- AudioOperational
- ModerationsOperational
- RealtimeOperational
- FilesOperational
- LoginOperational
- ConversationsDegraded performance
- LoginOperational
- ChatGPT WorkDegraded performance
- Codex in ChatGPT DesktopDegraded performance
- Compliance APIOperational
- SearchOperational
- File uploadsOperational
- Voice modeOperational
- GPTsOperational
- Image GenerationOperational
- Deep ResearchOperational
- AgentOperational
These are OpenAI's components, named and scored by OpenAI. We do not map them onto the functions we check above, because we have no way to know which of their components our checks touch. Two lists, side by side, neither one rewriting the other.
This is OpenAI's own claim about their service, shown separately from what we measured. We never change our own numbers based on what a vendor reports, and this reading never changes theirs. Read both.
OpenAI response time, by region
Is it OpenAI, or is it you?
Watching an outage above does not tell you whether your own connection or site is also having a problem. Enter your own URL for one live check, right now, no signup. This checks the site you enter, not OpenAI -- it cannot diagnose OpenAI's own systems.
What we measured, function by function
OpenAI answers on more than one surface, and they can fail apart. Each row is one surface, checked from the same regions. The verdict above follows the auth row, because that is the surface we treat as this service's own; a failure on any other row is a partial outage, not a full one.
- WebUS-East: OperationalUS-West: OperationalEurope: OperationalAsia-Pacific: OperationalUS-Central: OperationalCanada: OperationalUK: OperationalSoutheast Asia: OperationalAustralia: OperationalSouth America: Operational
- APIUS-East: OperationalUS-West: OperationalEurope: OperationalAsia-Pacific: OperationalUS-Central: OperationalCanada: OperationalUK: OperationalSoutheast Asia: OperationalAustralia: OperationalSouth America: Operational
- AuthprimaryUS-East: OperationalUS-West: OperationalEurope: OperationalAsia-Pacific: OperationalUS-Central: OperationalCanada: OperationalUK: OperationalSoutheast Asia: OperationalAustralia: OperationalSouth America: Operational
Get an email when we see OpenAI go down, and when it is back
Get told when we measure this
We will tell you when our probes stop reaching OpenAI, and again when they can reach it again. That is our own measurement from ten regions, not OpenAI's status page. No account, no card.
What we measured, and what people reported
- North AmericaOperationalUS-East, US-West, US-Central, Canada
- US-EastOperationallatency: 134msstatus code: 401checked: Oct 5, 2026, 8:15 PM UTC0 people in the last 5 minutesNot enough reports to sayOur US-East probes reach OpenAI normally. We do not have enough reports from US-East to say anything about what people there are seeing.
- US-WestOperationallatency: 113msstatus code: 401checked: Oct 5, 2026, 8:16 PM UTC0 people in the last 5 minutesNot enough reports to sayOur US-West probes reach OpenAI normally. We do not have enough reports from US-West to say anything about what people there are seeing.
- US-CentralOperationallatency: 118msstatus code: 401checked: Oct 5, 2026, 8:13 PM UTC0 people in the last 5 minutesNot enough reports to sayOur US-Central probes reach OpenAI normally. We do not have enough reports from US-Central to say anything about what people there are seeing.
- CanadaOperationallatency: 124msstatus code: 401checked: Oct 5, 2026, 8:14 PM UTC0 people in the last 5 minutesNot enough reports to sayOur Canada probes reach OpenAI normally. We do not have enough reports from Canada to say anything about what people there are seeing.
- South AmericaOperationallatency: 430msstatus code: 401checked: Oct 5, 2026, 8:13 PM UTC0 people in the last 5 minutesNot enough reports to sayOur South America probes reach OpenAI normally. We do not have enough reports from South America to say anything about what people there are seeing.
- EuropeOperationalEurope, UK
- EuropeOperationallatency: 256msstatus code: 401checked: Oct 5, 2026, 8:15 PM UTC0 people in the last 5 minutesNot enough reports to sayOur Europe probes reach OpenAI normally. We do not have enough reports from Europe to say anything about what people there are seeing.
- UKOperationallatency: 462msstatus code: 401checked: Oct 5, 2026, 8:15 PM UTC0 people in the last 5 minutesNot enough reports to sayOur UK probes reach OpenAI normally. We do not have enough reports from UK to say anything about what people there are seeing.
- Asia PacificOperationalAsia-Pacific, Southeast Asia, Australia
- Asia-PacificOperationallatency: 210msstatus code: 401checked: Oct 5, 2026, 8:16 PM UTC0 people in the last 5 minutesNot enough reports to sayOur Asia-Pacific probes reach OpenAI normally. We do not have enough reports from Asia-Pacific to say anything about what people there are seeing.
- Southeast AsiaOperationallatency: 239msstatus code: 401checked: Oct 5, 2026, 8:13 PM UTC0 people in the last 5 minutesNot enough reports to sayOur Southeast Asia probes reach OpenAI normally. We do not have enough reports from Southeast Asia to say anything about what people there are seeing.
- AustraliaOperationallatency: 575msstatus code: 401checked: Oct 5, 2026, 8:15 PM UTC0 people in the last 5 minutesNot enough reports to sayOur Australia probes reach OpenAI normally. We do not have enough reports from Australia to say anything about what people there are seeing.
Somewhere else or not sure: 0 people in the last 5 minutes
These are unverified reports from people using OpenAI, counted separately from what our own probes measure. Anyone can file one, no account is needed, and the region on a report is chosen by the person filing it. We never open, extend or close an outage in our record based on report volume: only our own measurements do that. We have not received any reports for OpenAI yet, which is not evidence that nothing is wrong.
Recent outages
measured entries are our own probes' readings; declared entries are the vendor's. How this check works
We have been measuring OpenAI since August 20, 2026. We show nothing from before that date, and we never add an outage we did not observe ourselves, from any source.
Need this in writing? Generate an SLA evidence report for OpenAI covering any window you pick, with our measurements and OpenAI's own claim at the time, side by side.
Full availability history for OpenAI: 90-day availability by region, month-by-month uptime, and every incident on record.
Monitor your own site the same way
This page shows what we see when we check this service, using the same ten regions, for a site you run. RealUptime Monitor watches your own servers from inside your network, and it starts free.
Start monitoring your own serversPut this badge in your README
Live, from this page's own readings.[](https://realuptime.io/outages/openai)<a href="https://realuptime.io/outages/openai"><img src="https://realuptime.io/outages/openai/badge.svg?days=30" alt="OpenAI uptime measured by RealUptime"></a>The badge shows the share of our probes that succeeded over the last 30 days, and the window is printed on it so the number always says what period it covers. If our readings go stale it says stale, and if we have no readings it says so: it will not show green on data we cannot vouch for. Free, no account, and it links back here. See how the badge works for every window and the honesty rules behind it.
What OpenAI's SLA implies, from our measurement
At or above your target
We measured none of downtime for OpenAI over the last 30 days, which is 100% uptime by our reading. A 99.9% target allows 43 minutes over that period.
By our measurement there is no shortfall to claim for this period.
This is computed from our own probe measurements, not from OpenAI's SLA instrumentation. Their contract is normally measured against their own systems, with their own exclusions and a claim window, so their figure and ours will differ. Use this to start the conversation and ask for their number. It is not legal advice and it is not a determination of what you are owed.