You might be concerned about Amazon Bedrock pricing details and specifications if you are considering using the product within the AWS platform.
Amazon Bedrock is a quick and easy way to build and implement Generative AI solutions with multiple options for leading foundation models on top of the fully managed capabilities of Amazon.
In this post, we will look at the pricing structure of Amazon Bedrock in detail, including the costs of different foundation models available on the platform.
Once you’ve signed up for AWS, you’re automatically enrolled in all its services, including Amazon Bedrock.
There’s no monthly subscription or annual fee to access the service, and you only pay for what you use. Your Bedrock costs are based on actual usage, ensuring that you get value for every penny spent.
As Bedrock is part of AWS, you can head to the Billing and Cost Management Dashboard in the AWS Billing and Cost Management console, where you can see a detailed breakdown of your usage and associated costs.
Amazon Bedrock Pricing Structure
When it comes to Amazon Bedrock, the charges are predominantly associated with running inferences on third-party foundation models. But what does that mean in practical terms?
Volume-Based Pricing
Costs are determined by the number of input and output tokens. Think of tokens as chunks of data or units of information that the model processes. Whether you’re feeding data into the model (input) or receiving processed information from it (output), you’re charged based on the volume of these tokens.
Provisioned Throughput
Some businesses might need a dedicated, faster processing capability, and for them, Amazon offers ‘provisioned throughput’. This is essentially a reserved capacity ensuring speedy and consistent processing. If you opt for this service, it comes with an additional charge.
Model-Specific Pricing
Since Amazon Bedrock offers a plethora of models from various AI companies, the pricing might vary depending on the model you choose. To get a clear picture of these costs, check the Model providers page in the Amazon Bedrock console. Each model version will have its associated pricing details listed.
Amazon Bedrock Pricing Details
When it comes to Amazon Bedrock, the charges are predominantly associated with running inferences on third-party foundation models.
AI21 Labs Models Pricing on Amazon Bedrock
On-Demand pricing for inference:
AI21 models | Price for 1000 input tokens | Price for 1000 output tokens |
Jurassic-2 Mid | $0.0125 | $0.0125 |
Jurassic-2 Ultra | $0.0188 | $0.0188 |
Currently, model customization (fine-tuning) and Provisioned Throughput are not supported for AI21 Lab models on Amazon Bedrock.
Amazon Titan Models Pricing on Bedrock
On-Demand pricing for model inference:
Amazon Titan models | Price for 1000 input tokens | Price for 1000 output tokens |
Titan Text – Lite | $0.0003 | $0.0004 |
Titan Text – Express | $0.0013 | $0.0017 |
Titan Embeddings | $0.0001 | n/a |
Pricing for model customization (fine-tuning):
Amazon Titan | Price to train 1000 tokens* | Price to store each custom model per month | Price to infer from a custom model for 1 model unit per hour (with no-commit Provisioned Throughput pricing) |
Titan Text – Lite | $0.0004 | $1.95 | $7.10 |
Titan Text – Express | $0.0008 | $1.95 | $20.50 |
*Total tokens trained = number of tokens in training data corpus x number of epochs
Provisioned Throughput pricing:
Amazon Titan models | Price per hour per model with no commitment* | Price per hour per model unit for 1-month commitment** | Price per hour for 6-month commitment** |
Titan Text– Lite | $7.10 | $6.40 | $5.10 |
Titan Text – Express | $20.50 | $18.40 | $14.80 |
Titan Embeddings | N/A | $6.40 | $5.10 |
*Custom model inference is limited to 1 model unit in the no-commit option.
**Includes inference for base and custom models. You can buy 2+ model units.
Anthropic AI Models Pricing on Amazon Bedrock
On-demand pricing:
Anthropic models | Price for 1000 input tokens | Price for 1000 output tokens |
Claude Instant | $0.00163 | $0.00551 |
Claude | $0.01102 | $0.03268 |
Provisioned Throughput pricing:
Anthropic models | Price per hour per model unit for 1-month commitment* | Price per hour for 6-month commitment* |
Claude Instant | $39.60 | $22.00 |
Claude | $63.00 | $35.00 |
*Includes inference for base and custom models.
Cohere Models Pricing on Amazon Bedrock
On-Demand pricing:
Cohere models | Price for 1000 input tokens | Price for 1000 output tokens |
Command | $0.0015 | $0.0020 |
Currently, model customization (fine-tuning) and Provisioned Throughput are not supported for Cohere models on Amazon Bedrock.
Stability AI Models Pricing on Amazon Bedrock
On-demand pricing: Image model offered by Stability are priced per image, depending on step count and image resolution.
Stability AI Model | Image Resolution | Standard quality (<51 steps) | Premium quality (>51 steps) |
SDXL0.8 | 512×512 or smaller | $0.018 per image | $0.036 per image |
Larger than 512×512 | $0.036 per image | $0.072 per image |
Provisioned Throughput pricing:
Stability AI model | Price per hour per model unit for 1-month commitment* | Price per hour for 6-month commitment* |
SDXL1.0 | $49.86 | $46.18 |
*Includes inference for base and custom models.
Currently, model customization (fine-tuning) is not supported for Stability AI models on Amazon Bedrock.
Amazon Bedrock’s pricing structure is transparent and usage-based, ensuring businesses only pay for the value they extract.