batch

Doubleword (async + batch inference) Tool

Run the same models you already use, but on async and batch tiers that trade latency for a large cost cut on workloads that don't need an instant reply: long-running agents, evaluations, and bulk jobs.