Amazon's mid-tier model. Good balance of quality and cost for RAG chatbots. Strong at following structured prompts and formatting rules.

Faster and cheaper than Nova Pro. Good for straightforward Q&A. Test whether the quality drop is noticeable for your visitors.

Text-only, lowest latency in the Nova family. Best for simple factual lookups. May struggle with complex formatting instructions.

Anthropic's latest flagship. Excellent at following nuanced system prompts, inline linking, and natural conversational tone. Most expensive option but often the best output quality.

Anthropic's fast, affordable model. Surprisingly capable for its price. Great candidate if Sonnet quality isn't needed for every query — compare to see the tradeoff.

Meta's latest open-weight model on Bedrock. Strong general-purpose performance with competitive quality. Worth testing for cost predictability.