Is OpenAI API down?
Loading the latest status…
Is OpenAI API down right now? Live status from OpenAI API’s status page, our own checks and user reports, updated every minute.
- Official status
- Reachable
- Answering prompts
- Reports in the last hour
Detailed data for this service is briefly unavailable. The status above comes from our latest check.
OpenAI API’s status page
- Chat Completions
- Responses
- Agents
- Decisions
- Fine-tuning
- Embeddings
- Images
- Batch
Show all 14 components
- Audio
- Moderations
- Realtime
- Files
- Login
- Live
Not working for me
Pick what’s broken. Your report shows up on the chart below within a few minutes.
Last 24 hours
- Operational
- Degraded
- Partial outage
- Major outage
- No data
- Reachability check (api.openai.com)
- Test prompt (OpenAI API)
By region
| Region | API api.openai.com | Test prompt OpenAI API |
|---|---|---|
| N. Virginia | ||
| Oregon | ||
| Frankfurt | ||
| Singapore | ||
| Tokyo | ||
| São Paulo |
Each row shows the latest result from that region. How we check
Last 90 days
We have checked OpenAI API ourselves since Oct 28, 2026. Earlier days come from its status page.
33 days with problems in the last 90 days.
Recent incidents
Incidents: 21 in the last 30 days and 37 in the last 90, lasting a median of 43 min. 10 problems our checks found never appeared on OpenAI API’s status page.
- OpenAI API not answering promptsPartial outage4 minFound by our checks, not on the status page
- GPT-image-2.5 and 2.6 Sunburst & Flare High error rateDegraded32 min
- Elevated errors affecting logins, sign-ups, and ChatGPT featuresDegraded9 min
- OpenAI API not answering promptsMajor outage16 minFound by our checks, not on the status page
- OpenAI API not answering promptsMajor outage6 minFound by our checks, not on the status page
- OpenAI API not answering promptsPartial outage2 minFound by our checks, not on the status page
- Elevated Errors On APIDegraded43 min
- OpenAI API not answering promptsPartial outage2 minFound by our checks, not on the status page
- OpenAI API not answering promptsPartial outage2 minFound by our checks, not on the status page
- OpenAI API not answering promptsPartial outage2 minFound by our checks, not on the status page
If OpenAI API isn’t working for you
The OpenAI API (platform.openai.com) gives developers programmatic access to GPT models for text, images, audio and embeddings.
Common OpenAI API errors
| Error | What it means | Usually | What to do |
|---|---|---|---|
401 - Incorrect API key provided | The API key in your request is wrong, revoked or belongs to another project. | Your side | Check the key and the Authorization header, or create a new key in the dashboard. Source |
429 - Rate limit reached for requests | You are sending more requests or tokens per minute than your limit allows. | Your side | Slow down, honor the Retry-After header and retry with exponential backoff. Source |
429 insufficient_quota (Credit balance exhausted) | Your organization has no prepaid credits left or hit a spend limit; retrying will not help. | Your side | Add credits or raise the spend limit in billing settings, then retry. Source |
500 - The server had an error while processing your request | Something failed on OpenAI's servers. | An outage | Retry after a short wait and check status.openai.com if it continues. Source |
503 - Model temporarily overloaded (server_is_overloaded) | The model you asked for does not have enough capacity right now. | Heavy load | Follow the Retry-After header and retry, or switch to another model for now. Source |
403 - Country, region, or territory not supported | The request comes from a location OpenAI does not serve. | Your side | Call the API from a supported country; a VPN exit in an unsupported region can also cause this. Source |
Is it just you?
- Read the status code first: 401, 403 and quota-type 429s come from your key, account or region, while 500 and 503 are on OpenAI's side.
- Run one minimal request with curl from a different machine or network to rule out your own proxy or firewall.
- Check the usage and limits pages in the dashboard to see whether you are near your rate limit or out of credits.
- Try a different model: a 503 on one model while others work is a capacity problem, not a full outage.
Alternatives to OpenAI API
Questions
A 429 means either you hit a rate limit (too many requests or tokens per minute) or your organization ran out of credits or hit a spend limit. Check error.code: rate limits clear if you back off, quota and billing errors only clear when you add credits or raise limits.
They share infrastructure but can fail separately. OpenAI's status page lists API and ChatGPT components individually, and this page sends real test requests to the API so you can see whether it is answering.
It is a 500 error from OpenAI's side. Retry after a short wait; if it persists across requests and models, check status.openai.com for an incident.
Sources: platform.openai.com, status.openai.com
Get alerted when OpenAI API breaks your app.
If your product depends on OpenAI API, your users feel its outages too. Point a Site Qwality monitor at the OpenAI API endpoint you use and at your own site, and we alert you on Slack, email, SMS or a webhook as soon as either one stops answering.