Skip to main content
llm.info
Back to Companies
Amazon logo

Cloud and AI company; creator of the Nova foundation models on AWS Bedrock.

4
Models
0
Apps
2
Tools
No apps linked to this company yet.

Amazon Bedrock

Amazon Bedrock is AWS's fully managed service for calling foundation models over an API, documented as providing "secure, enterprise-grade access to high-performing foundation models from leading AI companies". AWS states it supports 100+ foundation models from Amazon, Anthropic, DeepSeek, Moonshot AI, MiniMax, OpenAI and xAI, reachable through the bedrock-runtime endpoint via the Converse, Invoke, Responses, Chat Completions and Anthropic Messages APIs, alongside model customization (fine-tuning, continued pre-training, distillation). Use requires an AWS account plus IAM or Identity Center credentials, and capability coverage is region-scoped rather than global: the built-in Web Search tool is listed only for US East (N. Virginia), US East (Ohio) and US West (Oregon), and some models ship as "limited availability" to eligible customers only, as with OpenAI's Daybreak Red and Daybreak Blue. The service itself is closed-source and AWS does not enumerate the SDK languages on its Bedrock API setup page, saying only that AWS SDKs are "available for many popular programming languages".

Amazon Bedrock Guardrails

Amazon Bedrock Guardrails provides configurable safeguards to help you build safe generative AI applications, detecting and filtering undesirable content and protecting sensitive information in user inputs and model responses. AWS documents six safeguards: content filters (hate, insults, sexual, violence, misconduct and prompt attack), denied topics, word filters, sensitive information filters for PII and custom regex, contextual grounding checks for hallucination detection in RAG, and Automated Reasoning checks that validate responses against logical rules. Guardrails can be applied during an inference call by ID and version, or invoked on their own through the ApplyGuardrail API without calling a foundation model, and AWS states they work consistently across any foundation model whether you use models in Amazon Bedrock or self-hosted and third-party models such as OpenAI and Google Gemini. Trade-offs to weigh: it is a closed AWS service with no source, no self-hosted option and no published licence; screening explicitly excludes reasoning content blocks; supported Regions and models are limited to a documented list; and selectively evaluating only a section of an input prompt is available through the AWS SDK but not on the management console, including the Bedrock Playground and the Guardrails console.