How it's running right now.
The numbers below come from the gateway's public probe endpoints and are fetched when you open the page. Metrics that cannot be fetched remain unavailable rather than being replaced with placeholders.
Availability
and response performance.
The window is the last 24 hours, weighted by call volume. Latency is each model's average response time, success rate is the share of successful calls —— these are fields the backend actually records; we don't back out p99 from them or synthesize metrics that don't exist.
Upstream
health.
Reads current indicators from vendor status pages, subject to the gateway’s short cache. These colors describe vendor-reported status, not the availability you necessarily experience through EveryAPI.
For details about metering, routing, and the limits of public status data, see Technical transparency . If a request fails, contact support with its request ID and usage record.