Question

A legal-tech company is building a contract analysis assistant on Amazon Bedrock. Users upload single contracts that can be up to 250,000 tokens long, and the assistant must answer questions using the entire document in one prompt without chunking or retrieval. The team wants to minimize integration complexity by passing the full contract directly in the request. They are evaluating available foundation models. Which selection criterion should MOST directly drive their model choice?

Accepted Answer

Choose a model whose maximum context window is large enough to accept the full 250,000-token document plus the prompt and expected output

Answer

Choose the model with the lowest per-1,000-token input price, regardless of its context window size

Answer

Choose a model that supports provisioned throughput so the long documents are processed with dedicated capacity

Answer

Choose an embedding model with the highest output dimensionality to represent the full document