AWS Bedrock Pricing Calculator
Estimate Amazon Bedrock pricing by foundation model, input tokens, output tokens, monthly requests, prompt caching, batch share, discount assumptions, and non-token Bedrock charges. The calculator uses tracked Amazon Bedrock provider rows for planning and links to AWS for final regional billing verification.
Quick answer · Tracked pricing refreshed 2026-03-13 12:45:12
Amazon Bedrock cost starts with model inference, then changes with region, throughput mode, caching, batch jobs, guardrails, agents, knowledge bases, and data-transfer choices.
For on-demand text models, estimate input tokens multiplied by the input rate plus output tokens multiplied by the output rate. Use the calculator below for a monthly baseline, then add AWS services and Bedrock features that are billed separately.
Amazon Bedrock cost estimator
Model a monthly text-inference workload. Batch discount and regional adjustment are planning inputs because eligibility and rates depend on the selected model, region, and AWS pricing mode.
How to calculate AWS Bedrock pricing
- Choose the exact model and region. Bedrock model availability and pricing can differ by region and inference profile.
- Estimate tokens per request. Include system prompts, retrieved context, tool results, conversation history, and expected output.
- Multiply by monthly request volume. Separate interactive traffic from asynchronous jobs if they use different pricing modes.
- Apply cache and batch assumptions only when eligible. Cache reads apply to supported on-demand inference, while batch requests use batch rates without prompt caching. A discount should not be treated as universal across every model.
- Add non-model charges. Include guardrails, agents, knowledge bases, storage, vector search, data transfer, observability, and surrounding AWS services.
AWS Bedrock model pricing comparison
Prices below are tracked provider rows in USD per 1 million tokens. Verify the exact model ID, region, inference mode, and effective date on AWS before production budgeting.
| Model | Model provider | Input / 1M | Output / 1M | Context | Source |
|---|---|---|---|---|---|
| Claude Sonnet 4 | Anthropic | $3.00 | $15.00 | 200k | AWS pricing |
| Claude 3.5 Sonnet | Anthropic | $6.00 | $30.00 | 200k | AWS pricing |
| Nova Lite 1.0 | Amazon | $0.06 | $0.24 | 300k | AWS pricing |
| Nova 2 Lite | Amazon | $0.30 | $2.50 | 1.0M | AWS pricing |
| Nova Micro 1.0 | Amazon | $0.04 | $0.14 | 128k | AWS pricing |
| Claude 3 Haiku | Anthropic | $0.25 | $1.25 | 200k | AWS pricing |
| Nova Pro 1.0 | Amazon | $0.80 | $3.20 | 300k | AWS pricing |
| Claude 3.5 Haiku | Anthropic | $0.80 | $4.00 | 200k | AWS pricing |
| Claude Haiku 4.5 | Anthropic | $1.00 | $5.00 | 200k | AWS pricing |
| Palmyra X5 | Other | $0.60 | $6.00 | 1.0M | AWS pricing |
| Nova Premier 1.0 | Amazon | $2.50 | $12.50 | 1.0M | AWS pricing |
| Claude Opus 4.6 | Anthropic | $5.00 | $25.00 | 1.0M | AWS pricing |
| Claude Opus 4.5 | Anthropic | $5.00 | $25.00 | 200k | AWS pricing |
| Claude Opus 4.1 | Anthropic | $15.00 | $75.00 | 200k | AWS pricing |
| Claude Opus 4 | Anthropic | $15.00 | $75.00 | 200k | AWS pricing |
What changes an Amazon Bedrock bill?
Inference mode
On-demand, batch, prompt caching, cross-region inference, and Provisioned Throughput have different eligibility and billing mechanics.
Bedrock features
Guardrails, agents, knowledge bases, reranking, evaluation, customization, and data automation can add charges beyond model tokens.
AWS architecture
S3, OpenSearch Serverless, Aurora, Lambda, API Gateway, CloudWatch, NAT, and data transfer may materially change total cost.
AWS Bedrock cost examples
Customer support assistant
50,000 requests/month, 1,500 input tokens, and 400 output tokens on Claude Sonnet 4.
Token-only estimate before guardrails, retrieval, logging, and surrounding AWS infrastructure.
RAG knowledge assistant
250,000 requests/month, 5,000 input tokens, and 500 output tokens with retrieved context.
Add embeddings, vector storage/search, knowledge-base ingestion, and data transfer separately.
Calculator limitations
- Tracked provider rows may not include every model, region, marketplace offer, latency tier, or inference profile.
- The batch discount field is an assumption unless the selected model and job meet AWS batch requirements.
- Prompt caching is simplified to an on-demand cache-read share. Cache-write pricing, TTL behavior, minimum cacheable tokens, and cache misses can change the result, and prompt caching is not combined with batch inference.
- Provisioned Throughput and commitment terms require separate capacity planning.
- Use the official AWS Pricing Calculator for architecture-wide estimates and the AWS bill for actual usage.
Official AWS pricing sources
- Amazon Bedrock pricing — model inference and feature pricing.
- AWS Pricing Calculator — broader AWS architecture estimates.
- Amazon Bedrock documentation — service concepts, models, regions, and inference options.
AWS Bedrock pricing FAQ
How is AWS Bedrock pricing calculated?
For on-demand text inference, multiply input tokens by the selected model input rate and output tokens by its output rate, then add separately billed Bedrock features and AWS infrastructure.
Is Amazon Bedrock priced per request or per token?
Many text models are priced per input and output token, while image, video, embeddings, provisioned capacity, customization, guardrails, agents, knowledge bases, and other features can use different units.
Does AWS Bedrock batch inference reduce cost?
Some models and workloads can use discounted batch inference. Confirm model support, region, job requirements, and the current discount on the official AWS pricing page before applying it.
Does Bedrock pricing include vector database and RAG cost?
No. A RAG system may also incur embedding, ingestion, vector database, storage, search, Lambda, networking, and observability charges.
Is AWS Bedrock cheaper than calling a model provider directly?
It depends on the model, region, enterprise controls, networking, support, throughput mode, and surrounding architecture. Compare the exact provider row and total operating cost, not token price alone.
What is the difference between this tool and the AWS Pricing Calculator?
This page is a fast model-token estimator with tracked Bedrock rows. AWS Pricing Calculator is the better final tool for multi-service architecture estimates and region-specific AWS configurations.