The live state of every component, its uptime over the last 90 days, response times over the last 24 hours, and every incident and maintenance window since launch. All times are UTC.
Uptime counts major outages in full and partial outages at 60% of their duration; degraded performance and scheduled maintenance are not counted as downtime. Hover a marked day for its detail.
Latency is probed every minute from three vantage points outside our network; launch, notice and payment timings come from the platform itself. Each figure is the latest hour, the line is the last 24.
These are the numbers the deployment screen books against. They move through the day as jobs start and end; a zero means the pool is empty right now, not that the region is down.
Per-model availability is on the catalogue and on each model page. Need capacity guaranteed on a date? Reserve it.
No maintenance is scheduled. Windows are announced here, in the console and by e-mail to affected accounts, at least seven days before they start.
Completed — The maintenance is complete and every component is back to normal operation.
Completed — The maintenance is complete and every component is back to normal operation.
Completed — The maintenance is complete and every component is back to normal operation.
No incidents reported.
No incidents reported.
No incidents reported.
No incidents reported.
No incidents reported.
No incidents reported.
No incidents reported.
No incidents reported.
Resolved — This incident has been resolved.
No incidents reported.
Completed — The maintenance is complete and every component is back to normal operation.
No incidents reported.
No incidents reported.
No incidents reported.
Capacity is live, the catalogue is fresh and the console is up. Add credit and run something.
Billed per minute from the moment the instance is reachable. Minimum credit $30, no subscription.