No free plan. The cheapest way in is Claude Haiku (Anthropic) at $0.25 per 1M input tokens / $1.25 per 1M output tokens. Consider free alternatives in the ai category if budget is tight.
Amazon Bedrock provides access to hundreds of foundation models from leading AI companies, including Anthropic's Claude family, Meta's Llama, Mistral, Cohere, Stability AI, AI21 Labs, and Amazon's own Nova and Titan models. The model catalog expands regularly as new models are released, and built-in evaluation tools let you benchmark options against your specific use case. You can also import custom fine-tuned models, giving you flexibility to combine proprietary and third-party models in the same application.
Bedrock never stores or uses your data to train the underlying foundation models, and all data is encrypted in transit and at rest. Access is governed by AWS IAM identity-based policies, and traffic can be kept private using VPC endpoints. Bedrock is in scope for ISO, SOC, CSA STAR Level 2, GDPR, and FedRAMP High compliance, and is HIPAA eligible—making it one of the few generative AI platforms approved for regulated industries like healthcare, financial services, and government.
AgentCore is a composable agentic platform that lets you build, deploy, and operate agents at production scale using any framework or model without managing infrastructure. It includes nine independent services: Runtime, Gateway, Memory, Identity, Browser, Code Interpreter, Observability, Evaluations, and Policy. Bedrock Agents is the higher-level, guided agent-building experience, while AgentCore gives more advanced teams lower-level primitives to compose custom agentic architectures, including agents that interact with third-party tools and data sources.
Amazon Bedrock uses pay-as-you-go pricing based primarily on input and output tokens, with rates varying by model—so a call to a frontier model like Claude Opus costs more than a smaller model like Claude Haiku or Llama. Additional features including AgentCore services, Knowledge Bases, Guardrails, Model Distillation, and fine-tuning are billed separately. New AWS customers receive up to $200 in AWS credits to try AWS AI for free, and cost-optimization features like Intelligent Prompt Routing and prompt caching can cut inference bills by 30% or more.
Bedrock Guardrails can block up to 88% of harmful content via configurable content filters, denied topics, word filters, sensitive-information filters, and contextual grounding checks. For factual accuracy, Automated Reasoning checks use formal mathematical verification to identify correct model responses with up to 99% accuracy—a capability most competitors do not offer. Combined with Knowledge Bases for grounding responses in your proprietary data and Bedrock's evaluation tools, you can systematically reduce hallucinations and enforce safety policies in production.
See Amazon Bedrock plans and find the right tier for your needs.
See Pricing Plans →Still not sure? Read our full verdict →
Last verified March 2026