Deferred tier for autonomous agent scheduling (Preview): The deferred tier is available in Preview.

The deferred tier is available in Preview. The deferred tier automatically queues non-latency-sensitive background agent workloads and schedules them to execute during off-peak hours.

Key capabilities and benefits include:

• 50% token discount: Receive a 50% discount on consumed tokens for deferred workloads. • Reduced rate limiting: Mitigate resource exhaustion (429) errors and infrastructure pressure during long-running background tasks. • Supported agents:

Deep Research Agent: Pass service_tier="deferred" in the Python SDK or "service_tier": "deferred" in REST API interaction requests.

For more information, see Autonomous agent scheduling.