Amazon Bedrock
Serverless API access to foundation models (Anthropic, Meta, Amazon, and others) for building generative AI apps without managing infrastructure.
Verified against AWS documentation · 2026-08-27
Key facts
- 01One API, many models — pick a foundation model per request; no servers, clusters, or ML expertise required
- 02Knowledge Bases = managed RAG (ground answers in your own data); Agents = multi-step task orchestration; Guardrails = content filtering
- 03Customize models with fine-tuning on your data — the base model provider never sees your data, and your data never trains the shared model
- 04Inference modes: on-demand (per-token), batch (async, ~50% cheaper), provisioned throughput (dedicated capacity, hourly)
Exam triggers
Phrases and scenarios that point here.
- "Generative AI", "foundation models", "build a chatbot/summarizer with the least operational effort" → Bedrock
- "Use company documents to answer questions" (RAG) → Bedrock Knowledge Bases
- "Prevent harmful or off-topic model responses" → Bedrock Guardrails
- Fully managed / serverless GenAI vs. building and training your own model → Bedrock, not SageMaker
Not this service
- Build, train, and deploy your own ML models with full control → SageMaker
- Ready-made AI assistant for business users or developers, no building required → Amazon Q
Use cases
- 01Internal support chatbot that answers from your company wiki — Knowledge Bases handles the retrieval so answers cite real policy, not model guesses
- 02Nightly batch job summarizing thousands of customer support tickets into themes — batch inference at half the on-demand price since nobody's waiting on it
- 03An agent that takes 'rebook my flight' and orchestrates the lookup, cancellation, and rebooking API calls as separate steps
- 04Fine-tuning a base model on your product catalog's tone and terminology so generated descriptions sound like your brand, not generic marketing copy
- 05Guardrails on a public-facing assistant so it refuses off-topic or unsafe prompts before they ever reach the model
How it's charged
- 01On-demand: pay per input token and output token, rates set per model — model choice is the biggest cost lever
- 02Batch inference: same token metering at roughly 50% off on-demand, for async jobs that tolerate delay
- 03Provisioned throughput: dedicated model units billed hourly whether used or not, with 1- or 6-month commitments; required for fine-tuned/custom models
- 04Add-ons meter separately: Knowledge Bases (vector store + embeddings), Guardrails (per text unit checked), fine-tuning (per training token) — a single user request can trigger several billable operations
- 05No platform fee — Bedrock passes through model-provider rates and bills them via your AWS account
