Amazon Bedrock Verified
What is Amazon Bedrock?
Amazon Bedrock is AWS's fully managed platform for building generative AI applications with foundation models from Amazon, Anthropic, OpenAI, Meta, Mistral AI, DeepSeek, Google, xAI, Qwen, and other providers. It combines model inference with AWS identity, private networking, guardrails, evaluation, agents, and knowledge-base workflows, so teams can use managed models without operating inference infrastructure.
Amazon Bedrock API options
Use Converse or ConverseStream for a consistent multi-model message interface when the selected model supports it. Use InvokeModel when you need a provider-native request body or an operation not exposed through Converse. AWS also documents OpenAI-compatible and Anthropic-compatible API families for supported models.
- Verify API compatibility, model ID or inference profile, and Region before implementation.
- Converse uses bedrock-runtime and requires the same underlying model-invocation permissions as the inference operations.
- Treat streaming, tool use, guardrails, and model-specific request fields as capabilities to verify per model.
Amazon Bedrock model pricing
Representative on-demand US pricing, USD per 1M tokens, verified July 27, 2026 - check current regional rates at AWS Bedrock pricing.
| Model | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
| Claude Sonnet 5 | $2 promo; $3 standard | $10 promo; $15 standard |
| GPT-5.6 Sol | $5.50 | $33.00 |
| GPT-5.6 Terra | $2.75 | $16.50 |
| GPT-5.6 Luna | $1.10 | $6.60 |
| DeepSeek V3.2 | $0.62 | $1.85 |
| Gemma 4 31B | $0.14 | $0.40 |
| Mistral Large 3 | $0.50 | $1.50 |
| Grok 4.3 | $1.25 | $2.50 |
| Qwen3 Next 80B A3B | $0.15 | $1.20 |
Claude Sonnet 5 promotional pricing runs through August 31, 2026; standard pricing follows. Rates and availability vary by model, modality, region, inference tier, cache use, and cross-region routing. AWS also documents Standard, Flex, Priority, Reserved, and batch options; selected batch inference is 50% below on-demand pricing.
Key insights
Concrete technical or product signals.
- Bedrock now spans a broad multi-provider catalog, so model access should be verified by exact model ID, Region, and inference profile rather than provider name alone.
- AWS documents multiple service tiers and model-dependent batch options; cost comparisons should include routing tier, caching, and regional availability.
- Cloud identity, private networking, data location, and procurement are often more decisive than a single benchmark when comparing Bedrock with Azure OpenAI or Vertex AI.
Use cases
Where this shines in production.
- Enterprise copilots with IAM-scoped model access
- RAG and knowledge-base applications inside an AWS environment
- Multi-model routing across commercial and open-weight providers
- Agents and tool-using applications with AWS-native controls
- Text, image, speech, embedding, and multimodal production workloads
Limitations & trade-offs
What to watch for.
- Models, features, and quotas vary by AWS Region and inference profile.
- Provider-specific APIs and capabilities are not perfectly interchangeable even when a shared Bedrock API is available.
- Pricing comparisons must include token use, caching, inference tier, cross-region routing, and supporting Bedrock services.
Models referenced
Declared model dependencies or integrations.
Amazon Nova 2 Lite, Claude Sonnet 5, Claude Fable 5, Claude Mythos 5, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna, DeepSeek V3.2, Gemma 4 31B, Grok 4.3, Kimi K2.5, Llama 4 Maverick, Mistral Large 3, Qwen3 Coder Next, Cohere Embed v4
The list is representative, not exhaustive. AWS documents hundreds of models, and availability varies by Region, account access, inference profile, and release status. Verify the exact model ID before deployment.
Related prompts
Hand-picked or latest prompt templates.
Prompt
RAG Pipeline System Prompt Template
A production system-prompt template for retrieval-grounded answers with citation, access-control, and empty-retrieval handling rules.
Prompt
API Error Triage Workflow
A structured approach to identifying, categorizing, and resolving API errors in production systems.
Prompt
Model Evaluation Rubric for Production LLMs
A repeatable rubric for comparing production LLM candidates across quality, latency, cost, tool use, safety, and operational fit.
Prompt
Bedrock Converse API Integration Pattern
An implementation checklist for Bedrock Converse API integrations covering model IDs, retries, streaming, tool calls, IAM, and observability.
Prompt
Vector Embedding Pipeline for Enterprise RAG
A design template for enterprise embedding pipelines covering chunking, metadata, tenancy, indexing, refreshes, and retrieval evaluation.
Looking for a tighter match? Search the prompt library.
Amazon Bedrock vs Azure OpenAI - quick decision guide
| Dimension | Amazon Bedrock | Azure OpenAI |
|---|---|---|
| Model choice | Multi-provider catalog spanning Amazon, Anthropic, OpenAI, Meta, Mistral AI, DeepSeek, Google, xAI, Qwen, and others | Azure-managed access centered on OpenAI models |
| Cloud fit | AWS-native identity, networking, observability, and procurement | Azure-native identity, networking, observability, and procurement |
| Identity and private access | IAM, service control policies, and VPC endpoints | Microsoft Entra ID, Azure RBAC, and private endpoints |
| Best fit | AWS teams that want one managed route to several model providers | Azure teams standardizing on OpenAI models and Microsoft services |
| Capacity options | On-demand plus model-dependent Flex, Priority, Reserved, batch, and provisioned options | Usage-based deployment plus provisioned throughput options |
Choose Amazon Bedrock when AWS is your operating environment or provider choice is a core requirement. Choose Azure OpenAI when the application is Azure-native and specifically centers on OpenAI models. Compare regional model access, data residency, quotas, and measured workload cost before committing.
Amazon Bedrock FAQ
What is Amazon Bedrock?
Amazon Bedrock is a fully managed AWS service for building generative AI applications with foundation models and AWS-native controls. It provides managed model inference plus services for agents, knowledge bases, guardrails, evaluation, and related workflows.
Is Amazon Bedrock an AI model?
No. Amazon Bedrock is a managed platform and model gateway. It provides access to models from Amazon and third-party providers through AWS services and APIs.
Which AI models are available in Amazon Bedrock?
The catalog includes models from Amazon, Anthropic, OpenAI, Meta, Mistral AI, DeepSeek, Google, xAI, Qwen, Cohere, and other providers. Exact model IDs and availability differ by AWS Region and can change over time.
How does Amazon Bedrock pricing work?
Pricing depends on the provider, model, modality, token volume, AWS Region, and inference tier. AWS offers model-dependent on-demand, batch, Flex, Priority, Reserved, and provisioned options, so production estimates should use the exact model ID and deployment Region.
What is the difference between Amazon Bedrock and Azure OpenAI?
Amazon Bedrock is AWS-native and exposes a broad multi-provider model catalog. Azure OpenAI is Azure-native and centers on managed OpenAI models. The practical choice usually follows cloud identity, networking, data location, procurement, model requirements, and measured cost.
Can Amazon Bedrock be accessed privately from a VPC?
Yes. AWS supports private connectivity patterns for Bedrock through VPC interface endpoints powered by AWS PrivateLink. Teams should verify endpoint support for the specific Bedrock APIs and Region they plan to use.
Official sources
- AWS: Bedrock API compatibility
Current API families and model-level compatibility.
- AWS: Converse API with Boto3
Official Python setup and Converse request example.
- AWS: Models at a glance
Current provider and model catalog.
- AWS: Amazon Bedrock pricing
Official model and service pricing.
Related
Comparisons, platforms, and models teams often view next.