Best-effort quota only runs after provisioned is exhausted - and has no SLA

Do not size your pipeline on the best-effort numbers - that quota only kicks in after provisioned is exhausted, and it carries no SLA, so bursts can still throttle or degrade. If you need reliable throughput for the Pro model at 30 pages/min provisioned, request a quota increase or a capacity reservation instead of hoping best effort covers the gap.

Context: Google Cloud Document AI quotas docs: online processing for GenAI-powered processors has two tiers - provisioned (120 pages/min for Flash-based custom extractor, 30 for Pro) and best effort (120, or 60 for Pro). Best effort is only used once the provisioned quota is exhausted, and there is no SLA on it.

Source: Google Cloud Document AI quotas docs: online processing for GenAI-powered processors has two tiers - provisioned (120 pages/min for Flash-based custom extractor, 30 for Pro) and best effort (120, or 60 for Pro). Best effort is only used once the provisioned quota is exhausted, and there is no SLA on it.

Verified resolution: Do not size your pipeline on the best-effort numbers - that quota only kicks in after provisioned is exhausted, and it carries no SLA, so bursts can still throttle or degrade. If you need reliable throughput for the Pro model at 30 pages/min provisioned, request a quota increase or a capacity reservation instead of hoping best effort covers the gap.

Source URL: http://cloud.google.com/document-ai/quotas