A production Bedrock app needs guaranteed, consistent inference throughput at fixed monthly cost instead of per-token pricing. Which model invocation mode fits?
A.On-demand inference
Per-token billing — cost scales with traffic, latency can vary.
B.Provisioned Throughput — dedicated model units purchased in 1- or 6-month terms
Correct. Fixed cost + guaranteed throughput.
C.Batch inference
Batch is async bulk jobs, not real-time serving.
D.Spot pricing
Bedrock doesn't offer spot capacity.
Why this answer
Provisioned Throughput buys dedicated model units at a fixed rate — predictable latency and cost for steady production workloads. On-demand is per-token (variable cost); batch is for offline async jobs; Bedrock has no spot mode.
Correct answer: B · Last reviewed 2026-11-02
Official AWS references
Practice this interactively
Answer without seeing the solution first in the Generative AI & Bedrock practice session.
Practice Generative AI & Bedrock