Skip to main content
Sailboxes charge only for the CPU, memory, and disk a Sailbox actually uses while running. If a Sailbox with a 4 vCPU maximum is using 0.01 vCPU, you are charged for 0.01 vCPU. You are not charged while a Sailbox is sleeping, paused, checkpointing, cold-starting, or blocked on a Sail inference request. Every Sailbox creation also has a one-time $0.01 charge. This billing model works because Sailboxes run on spot instances and live-migrate between hosts during periods of contention. Agentic workloads spend most of their time blocked on I/O, so paying for observed usage instead of reserved capacity makes long-horizon agents far cheaper to run.
DimensionPrice
Used vCPU/hour (Graviton4)$0.03
Used RAM (GB)/hour$0.008
Used NVMe disk (GB)/hour$0.0007
Sailbox creation$0.01