Rate Limits
Default request rate limits for Judgment's platform APIs.
Judgment enforces per-organization rate limits on a few high-traffic request paths. Each limit is a token bucket: your organization has a maximum burst capacity, and tokens refill at a steady rate. A request is allowed if a token is available, and rejected once the bucket is empty.
| Scope | What it covers | Default limit |
|---|---|---|
| Evaluation queue | Dispatching online evaluations | 100 evaluations / min |
| Offline tests | Evaluations created by Offline Tests | 500 evaluations / min |
| Public queries | Public JQL requests made via the SDK/API | 40 requests / min |
| Dashboard queries | JQL requests made by the dashboards UI | 50 requests / min |
For the evaluation queue and offline tests scopes, each request consumes one token per evaluation run it schedules (examples × judges), not one token per request — a single API call that fans out into many evaluation runs can consume many tokens at once.