Is the OpenAI API down?
The API rows from status.openai.com are below. During an outage, requests fail with HTTP 503 "server_is_overloaded", HTTP 500 or HTTP 502. Retry with backoff and honor Retry-After, switch model in your agent, or reach the same GPT model on Azure.
Live OpenAI API status
Checked
All systems operational
- Chat CompletionsOperational
- ResponsesOperational
- RealtimeOperational
- BatchOperational
- EmbeddingsOperational
- Last incident: Degraded performance in GPT-6.1 Sol Codex and Work moderesolved Oct 13
Source: status.openai.com. Read on our server about every 2 minutes.
Which error is OpenAI, and which is you
| What you see | What it means |
|---|---|
HTTP 503, server_is_overloaded | The model has no spare capacity right now. Follow Retry-After. See the 503 fix |
| HTTP 500 | "The server had an error while processing your request": an error at OpenAI. Retry after a short wait, then check the status page |
| HTTP 502, Bad gateway | A transient error on OpenAI's side. Retry |
HTTP 429, rate_limit_exceeded | "Rate limit reached for requests": your tier's own limit. See the 429 fix |
HTTP 429, insufficient_quota | Billing. See the quota fix |
Five minutes into an OpenAI outage
- Read the Chat Completions and Responses rows above. ChatGPT has its own rows and its own incidents.
- Retry with exponential backoff and honor the Retry-After header when OpenAI sends one.
- If one model is overloaded, switch to another GPT model or to another vendor's model with the steps below.
- Or keep the model and change the provider: Azure runs OpenAI models on its own infrastructure, with its own status page.
How to switch model in your agent
| Agent | Steps |
|---|---|
| Codex CLI | Type /model and pick another model. For another provider, run standardcompute launch codex or add a [model_providers] entry to ~/.codex/config.toml |
| Cline, Roo Code, Kilo Code | Open the extension's settings, change API Provider or Model, and save. The next request uses it |
| OpenCode | Type /models and pick a model from another provider, or start with opencode -m provider/model |
| Aider | Type /model followed by another model name; the chat history stays |
| Your own code | Catch 5xx and 529, retry with exponential backoff, then send the request to a fallback model |
What Standard Compute does when OpenAI fails
- Every request goes through OpenRouter. GPT requests go to OpenAI first; if OpenAI fails, OpenRouter can send the request to another provider of the same GPT model, such as Azure.
- Each failed call is retried twice on a 429, a 5xx or a network error, after 0.5 and then 1 second, or after the provider's Retry-After up to 8 seconds. If every attempt fails, your agent gets HTTP 503 with "The model is temporarily unavailable. Please retry shortly."
- A specific model ID always stays that model; failover only changes the provider. Only the auto-router, the model
standardcomputewith no model selection, can switch to a different model. - A key restricted to Europe or the US fails over only between the endpoints approved for that region.
The catch
Failover needs a provider that still works. When a GPT model fails at OpenAI and Azure together, Standard Compute cannot serve it either; switch to another model on the same key, such as anthropic/claude-sonnet-5.5 or deepseek/deepseek-v4.1-flash, or wait.
Where Standard Compute fits
Standard Compute is an OpenAI- and Anthropic-compatible LLM gateway for coding agents at https://api.stdcmpt.com/v1, with monthly plans from $19 (USD) and a fixed compute budget. One key serves Claude, GPT, Gemini, DeepSeek and GLM models, so moving off a struggling model during an outage takes one changed model ID.
FAQ
Is the OpenAI API down or is it just me?
Check the Chat Completions and Responses rows on status.openai.com. If they are operational and requests still fail, read the error: 401 is your key, 429 with rate_limit_exceeded is your tier's limit, 429 insufficient_quota is billing, and 503 server_is_overloaded is OpenAI's capacity.
Is ChatGPT being down the same as the OpenAI API being down?
No. ChatGPT and the API are separate components on status.openai.com, and an incident can hit one without the other. For your code or coding agent, read the API rows.
Does Azure OpenAI keep working when OpenAI is down?
It can, because Azure runs OpenAI models on its own infrastructure with its own status page. It can still have incidents of its own, and new models sometimes arrive on Azure later.
Can a gateway keep GPT requests working during an OpenAI outage?
Only while another provider of the same model works. On Standard Compute a GPT request goes to OpenAI first and can fail over to another provider of that model such as Azure, with two retries, and returns HTTP 503 if every attempt fails. A specific model ID always stays that model.
Does a gateway add latency?
Yes, a little. Production logs for October 13, 2026 put the gateway's own work before a request leaves (key, budget and policy checks) at a median of about 0.6 seconds, and about 1.1 seconds when a request waits for the auto-router's model choice. Send a specific model ID to skip the router.