Try it yourself

Explore

Run any question through the same four retrieval modes the Benchmark uses — Flat, Hierarchical, LLM-only, and RAG — and see the latency, cost, and answer quality tradeoff for yourself, one query at a time.

docs indexed: 116 docs indexed
retrieval modes: 4 retrieval modes
avg latency (session): — avg latency (session)
LLM calls (session): 0 LLM calls (session)

Ask a question

Retrieval mode

Flat (Lexical)Ranks every chunk in the index directly against the query — a single global BM25-style pass.HierarchicalNarrows document → section → chunk before ranking, trading a little latency for more precise context.LLM-onlyNo retrieval step — answers from the model's parametric memory alone. Fast, but ungrounded.RAGRetrieves grounding context first, then generates an answer from it — the full retrieval-augmented pipeline.

Live index: 116 curated AWS documentation pages, 4 retrieval modes, real Bedrock calls for LLM and RAG.

Enter a question above, or pick an example query, to see results for the selected mode.