Deferred tier for autonomous agent scheduling (Preview)

Deferred tier for autonomous agent scheduling (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.