# Status page

https://modelportal.ai/docs/status-page

The [status page](/status) shows, per model, whether it is up now, along with availability, time to first token and throughput over the last 4 hours, 24 hours and 7 days.

## Where the data comes from

- Real calls: only platform-side failures count against availability; your own request errors (invalid parameters, refused content, cancellations and so on) do not.
- Scheduled probes: every 5 minutes the platform sends one tiny request to each callable text model. Probes are not billed to anyone.
- Periods with too few calls are marked as low-sample, which does not mean the model is down.

## Incidents and maintenance

Incidents and planned maintenance are posted as announcements with start and end times and the affected endpoints, shown both on the status page and on the matching model pages. Past entries are in [Announcements](/announcements).

## Model statuses

What Available, Degraded, Maintenance, Retiring and Retired mean on a model page is covered in [Model capabilities](/docs/capabilities).