Bedrock Pricing Calculator - AI Model & Tool Cost Estimator
Amazon Bedrock pricing calculator to estimate costs for foundation models (Claude, Llama, Mistral, Nova) and AI tools (Flows, Guardrails, Prompt Routing). Compare pricing across providers and AWS regions.
Free price alerts · Platform launching early 2027
Get an email when Bedrock prices change
CloudBurn re-checks AWS prices every week and emails you when a price behind this calculator changes.
30% founding-member discount
Only waitlist members get it. Join now and keep 30% off for as long as you keep your plan.
Alert emails only arrive when a price changes; the waitlist sends two a month at most. Unsubscribe anytime. Privacy policy
Tick the box to also save your spot on CloudBurn Platform:
- Price your whole stack in minutesThe AWS Simple Pricing Calculator puts Bedrock and the rest of your AWS services in one estimate, ready to export.
- Know the cost before you deployThe GitHub App estimates what your Terraform, CDK or CloudFormation changes will cost before they ship.
- Find the waste, then fix itScans of your live AWS accounts show what is costing money for nothing, with clear steps to fix each finding.
Frequently Asked Questions
How does Amazon Bedrock pricing work?
Amazon Bedrock pricing is based on usage with no upfront commitments. For text models, you pay per 1,000 input and output tokens. Image generation models charge per image, while video models charge per second. Tool features like Guardrails and Flows have separate per-use pricing. Prices vary by model provider and AWS region.
What are the costs for different Bedrock foundation models?
Bedrock model pricing varies significantly by provider and model size. For example, in us-east-1:
- Claude 3.5 Sonnet: $0.003/1K input, $0.015/1K output tokens
- Llama 3 70B: $0.00099/1K input, $0.00099/1K output tokens
- Amazon Nova Pro: $0.0008/1K input, $0.0032/1K output tokens
- Mistral Large: $0.004/1K input, $0.012/1K output tokens
Use the calculator above to see current pricing for all models in your preferred region.
How much does Bedrock Guardrails cost?
Bedrock Guardrails pricing depends on the policy type. Content filters cost $0.15/1K text units, topic filters cost $0.15/1K units, sensitive information detection (PII) costs $0.10/1K units, and contextual grounding costs $0.10/1K units. Word filters are free. Image content filtering costs $0.75/1K images. A text unit is approximately 1,000 characters.
What is Bedrock Flows pricing?
Bedrock Flows charges $0.035 per 1,000 node transitions. A node transition occurs each time your workflow moves from one step to another. This is in addition to any model inference costs. Flows is ideal for building multi-step AI workflows with visual orchestration.
How does Intelligent Prompt Routing pricing work?
Intelligent Prompt Routing costs $1.00 per 1,000 routing requests. It automatically selects the optimal model based on your prompt, balancing cost and quality. The routing cost is separate from the actual model inference cost. This feature can reduce overall costs by selecting cheaper models when appropriate.
What Bedrock pricing tiers are available?
Amazon Bedrock offers several pricing tiers:
- Standard: Pay-as-you-go with no commitments (shown in this calculator)
- Priority: Premium pricing for lower latency
- Flex: Discounted pricing for non-urgent, flexible workloads
- Reserved: Committed capacity with significant discounts for consistent usage
- Batch: 50% off standard pricing for asynchronous processing
Prompt Caching offers up to 90% savings on cached tokens for supported models.
How can I reduce Amazon Bedrock costs?
Key strategies to optimize Bedrock costs:
- Use prompt caching: Up to 90% savings on repeated prompts
- Choose smaller models: Use Haiku or Nova Lite for simpler tasks
- Batch inference: 50% discount for non-time-sensitive workloads
- Intelligent Prompt Routing: Auto-select cost-effective models
- Optimize prompts: Shorter, focused prompts reduce token costs
For more cost optimization strategies, use our Lambda pricing calculator to model serverless AI workloads.
Related tools