We’re excited to announce that Claude Fable 5 is now accessible via Amazon Bedrock and the Claude Platform on AWS. This cutting-edge AI model brings Mythos-tier performance to a wider audience, while maintaining robust safety measures for responsible deployment. Fable 5 achieves near the highest results across all benchmarks tested and excels in software engineering tasks, knowledge work, and visual understanding – making it perfect for handling demanding, time-intensive projects.
Running Claude Fable 5 through Amazon Bedrock lets you build solutions within your existing AWS infrastructure and seamlessly expand your inference workloads. Alternatively, you can access the model through the Claude Platform on AWS, providing you with Anthropic’s original platform experience.
As noted by Anthropic, Claude Fable 5 represents a major breakthrough in what you can achieve with AI models. Here’s what makes this model truly stand out:
- Extended autonomous processing capabilities — Claude Fable 5 can manage intricate tasks that earlier models couldn’t handle for long periods, performing coding and knowledge work operations continuously without requiring human input.
- Enhanced visual understanding — The model can accurately interpret diagrams, charts, and tables embedded in documents and PDFs. This enables sophisticated research and analysis across industries including finance, legal, analytics, architecture, and gaming. For coding tasks, it faithfully implements design specs and leverages visual insights to evaluate its own output against objectives.
- Built-in self-checking mechanisms — The model independently improves its capabilities through learning, creates its own testing frameworks, and validates its own work.
Claude Fable 5 incorporates protective measures that intentionally limit its performance in areas where misuse risks are higher. Requests involving cybersecurity, biology, chemistry, and health topics are automatically redirected to Opus 4.8 for responses. By developing stronger safety protocols, Anthropic has been able to make nearly all of Claude Fable 5’s advanced capabilities available to users. The unrestricted version of this model is called Claude Mythos 5, which will only be offered to a select group of approved customers.
Claude Fable 5 in practice
You can leverage Claude Fable 5 through both Amazon Bedrock and the Claude Platform on AWS. This article focuses on how to access and use the model on Amazon Bedrock. For instructions on using the Claude Platform on AWS, please refer to the official documentation.
To begin working with Amazon Bedrock, you can now access the model programmatically using the Anthropic Messages API to connect to the bedrock-runtime or bedrock-mantle endpoints through the Anthropic SDK. You can also continue using the Invoke and Converse APIs on bedrock-runtime via the AWS Command Line Interface (AWS CLI) and AWS SDK.
Before you can use the Claude Fable 5 model, you need to enable data sharing by configuring the Data Retention API and setting provider_data_share. Please note that there is no console interface available for this setting at launch.
Below is a sample script to configure data retention for the bedrock-mantle engine:
curl -X PUT
-H "x-api-key: "
-H "Content-Type: application/json"
-d '{ "mode": "provider_data_share" }' If you prefer to use the bedrock-runtime engine, execute this sample script:
curl -X PUT
-H "Authorization: Bearer "
-H "Content-Type: application/json"
-d '{ "mode": "provider_data_share" }' This setting permits Amazon Bedrock to store and share your inference data with model providers according to their requirements. Anthropic mandates 30-day retention of inputs and outputs, along with human review processes. For additional details, please visit the Amazon Bedrock abuse detection page.
Let’s begin with the Anthropic SDK for Python using the Messages API on the bedrock-mantle endpoint. First, install the Anthropic SDK:
pip install anthropicHere’s a sample Python script to invoke the Claude Fable 5 model:
import anthropic
client = anthropic.Anthropic(
base_url="
api_key=
)
message = client.messages.create(
model="anthropic.claude-fable-5",
max_tokens=4096,
messages=[
{ "role": "user",
"content": "Design a distributed architecture on AWS in Python that should support 100k requests per second across multiple geographic regions",
},
],
)
print(message.content[0].text) For more information, explore the Anthropic Messages API code examples and notebook samples covering various use cases and multiple programming languages.
You can also interact with Claude Fable 5 directly in the Bedrock console. Simply select Claude Fable 5 in the Playground section to test it out.
Additionally, you can use Claude Fable 5 with the Invoke API and Converse API on the bedrock-runtime endpoint. Here’s an example demonstrating how to call the Converse API for a unified multi-model experience using the AWS SDK for Python (Boto3):
import boto3
bedrock_runtime = boto3.client("bedrock-runtime", region_name="us-east-1")
response = bedrock_runtime.converse(
modelId="global.anthropic.claude-fable-5",
messages=[
{
"role": "user",
"content": [
{
"text": "Design a distributed architecture on AWS in Python that should support 100k requests per second across multiple geographic regions."
}
]
}
],
inferenceConfig={
"maxTokens": 4096
}
)
print(response["output"]["message"]["content"][0]["text"]) For further details, check out the code examples demonstrating how to use Amazon Bedrock Runtime with AWS SDKs.
Key information to keep in mind
Here are some important technical details that you should be aware of:
- Model availability — Access to Claude Fable 5 is being gradually rolled out to all AWS accounts. If your account doesn’t have access yet, it will be granted soon based on your Bedrock usage patterns. To expedite access to this model, reach out to your regular AWS Support contact.
- Cost structure — When a potentially harmful request is redirected to Opus 4.8 instead of Fable 5, you’ll only be charged Opus rates. If a request is interrupted mid-conversation, initial tokens are billed at Fable rates while subsequent tokens are charged at Opus rates. For complete pricing details, visit the Amazon Bedrock pricing page.
- Data retention policy — For Fable 5, Mythos 5, and future models on Bedrock with comparable or higher capability levels, Anthropic requires 30-day retention for all traffic on Mythos-class models. This limited retention period helps Anthropic identify patterns of misuse that aren’t apparent from individual interactions. Once you enable data retention, your data will move outside AWS’s data and security boundaries.
- Claude Mythos 5 on Bedrock (Limited Preview) — You can also access Anthropic’s most powerful model for cybersecurity and life sciences applications, including vulnerability identification, drug development, and biodefense screening. Access is currently restricted due to the dual-use potential of these domains. For more information, visit the model card documentation.
Available now
Anthropic’s Claude Fable 5 model is available today on Amazon Bedrock in the US East (N. Virginia) and Europe (Stockholm) Regions; refer to the complete list of Regions for upcoming availability updates. Claude Fable 5 is also accessible on the Claude Platform on AWS across North America, South America, Europe, and Asia Pacific.
Try out Claude Fable 5 using the Amazon Bedrock APIs or through the Claude Platform on AWS, and share your feedback on AWS re:Post for Amazon Bedrock or through your regular AWS Support channels.
— Channy
Updated on June 9, 2026 — 1) Updated the console screenshot. You can now use the console with the bedrock-runtime engine. Console support for bedrock-mantle is coming soon. 2) Corrected the model ID in the sample code, 3) Fixed the proper provider_data_share parameter, 4) Added a data retention configuration script for the bedrock-runtime engine.
