AI API status — is the API down?
Live status of the AI APIs developers build on: OpenAI, Claude, Gemini, Grok, DeepSeek, Mistral, OpenRouter, Hugging Face, Replicate and ElevenLabs. Status comes from official feeds where they exist; latency is our own measured check, updated every minute.
| API | Status | Latency | 90-day uptime |
|---|---|---|---|
| ChatGPTOpenAI · checked 5s ago · official feed | Operational | 650ms | 98.89% |
| ClaudeAnthropic · checked 5s ago · official feed | Operational | 564ms | 96.99% |
| GeminiGoogle · checked 5s ago · official feed | Operational | 1212ms | 99.96% |
| GrokxAI · checked 5s ago · our check | Operational | 653ms | n/a |
| DeepSeekDeepSeek · checked 5s ago · official feed | Operational | 1009ms | 99.87% |
| MistralMistral AI · checked 8s ago · official feed | Operational | 361ms | 100% |
| OpenRouterOpenRouter · checked 5s ago · official feed | Operational | 1218ms | n/a |
| Hugging FaceHugging Face · checked 5s ago · official feed | Operational | 554ms | 99.94% |
| ReplicateReplicate · checked 5s ago · our check | Operational | 553ms | n/a |
| ElevenLabsElevenLabs · checked 5s ago · official feed | Operational | 630ms | 98.76% |
Uptime is computed from each vendor's published incident history (major = full weight, partial = 50%, degraded = 25%). “n/a” means the vendor publishes no history.
API outage FAQs
Is the OpenAI API down?
See the live table above — the ChatGPT / OpenAI row reflects OpenAI's official status feed, refreshed every minute. API-specific incidents (elevated errors, latency) are posted there before anywhere else.
How do I tell an API outage from my own bug?
If the status above is operational but your calls fail: 401 means an invalid key, 429 means rate limits, 529/503 usually means provider overload — retry with backoff. Only 5xx across multiple models plus a status-page incident means a real outage.
Which AI API is most reliable?
Compare the 90-day uptime column above — it's computed from each vendor's own published incident history with identical weighting, so the numbers are comparable. No vendor publishes latency SLAs on these pages; the ms column is our own measured check.