SERVICE STATUS
Checking service status…
Connecting to Instatus
Get an email when we post an update.
Stop issue emails?
This only stops updates for the issue linked in your email.
Current issues
Loading…
Services
Loading…
Model performance & availability
Models
No matching models
Period
Loading recorded checks…
Throughput
Model check history
Passed checks Failed checks Missing checks
These are sampled model checks, not the success rate of all customer requests. Missing checks are unknown and excluded from the pass rate.
What counts as a passed, failed or missing check?
Every 10 minutes, the monitor sends a short streaming chat request to each model using its own API key. A failed attempt is retried once.
- Passed
- The probe received at least one content or reasoning token without a reported API error. A retry can turn an initial failure into a pass. A stream that disconnects or times out after its first token can also pass; completion of the full response is not verified.
- Failed
- Both attempts failed: for example, an HTTP error, an error returned in the stream, a connection failure, no tokens, or the 60-second request timeout before any token arrived.
- Missing / unknown
- No usable result is available for an elapsed 10-minute slot. The monitor may not have recorded it, or the history file may be unavailable. This does not establish whether the service was up or down. Future slots are excluded.
Checks cover this monitor’s request path and account. They may miss issues affecting other accounts, routes or requests. The performance lines are averaged; the pass rate uses the underlying check results.