A media company runs a nightly job that summarizes 500,000 archived news articles using a foundation model on Amazon Bedrock. The summaries are stored in S3 and are not needed until the next business day. The team currently calls the InvokeModel API in a loop and is frustrated by per-request throttling and high on-demand costs. They want the most cost-effective approach that avoids managing throughput capacity for this large, latency-tolerant workload. Which approach should they use?