A media company runs a steady, high-volume summarization job through a foundation model around the clock and needs predictable latency at a lower unit cost. Which Amazon Bedrock option should it evaluate?
Show answer and explanation
Answer: C
Provisioned Throughput reserves a fixed amount of inference capacity for a specific model. That gives steady latency under sustained load and a discounted rate in exchange for a commitment term, with longer terms costing less per hour. Charges continue until the provisioning is deleted, so it fits predictable demand. On-demand billing suits spiky or small workloads because it reserves nothing and latency can vary under load. A free sandbox is for experimentation, and cross-Region routing improves availability rather than offering reserved capacity.
Why the other options are wrong
- A. On-demand pricing charges per token with no reservation, which suits variable traffic, not a steady, always-on workload.
- B. Free quotas are meant for evaluation and cannot support a production workload at this volume.
- D. Cross-Region routing helps absorb traffic and improve availability, but it does not reserve capacity or lower the per-hour rate.
Exam tip. Steady, high volume favors provisioned throughput; spiky or small workloads favor on-demand.