Anthropic Releases Claude 3.5 Sonnet: New Capabilities on Amazon Bedrock

Anthropic has officially released its latest AI model, Claude 3.5 Sonnet, now available to developers and enterprises through Amazon Bedrock and the Claude Platform on AWS. The model offers significant improvements in software engineering, vision-based reasoning, and long-form content generation compared to previous iterations. According to official company statements, Claude 3.5 Sonnet is designed to handle complex, asynchronous tasks while maintaining high performance across standardized industry benchmarks.

What distinguishes Claude 3.5 Sonnet from previous versions?

The primary shift in Claude 3.5 Sonnet is its ability to execute long-running, multi-step workflows without constant human intervention. Anthropic reports that the model excels at coding tasks, where it can implement designs and utilize vision capabilities to critique its own output against specific project goals. Unlike its predecessors, the model features advanced vision processing, allowing it to interpret charts, graphs, and complex diagrams embedded within PDFs or other documents. This capability is specifically targeted at industries such as finance, architecture, and legal analysis, where document-heavy workflows are standard.

How to access the model on Amazon Bedrock

Developers can integrate Claude 3.5 Sonnet into their existing AWS environments using the Amazon Bedrock API. To begin, users must ensure their account has the necessary permissions and that they have opted into data sharing via the Data Retention API. Anthropic requires a 30-day retention period for inputs and outputs on this model to facilitate misuse detection, a policy that differentiates it from models with lower capability tiers. Users can interact with the model programmatically using the AWS SDK or the AWS Command Line Interface (CLI) by configuring the provider_data_share mode.

Safety protocols and model routing

Anthropic has implemented a tiered safety architecture for its latest release. To mitigate risks in sensitive domains—such as cybersecurity, advanced chemistry, or biological research—the system utilizes a routing mechanism. When the model detects a prompt that triggers specific safety thresholds, it automatically redirects the query to the Claude 3 Opus model for processing. This ensures that users receive high-fidelity responses while maintaining robust guardrails. Users are billed according to the model used for each specific request; requests routed to Opus are charged at standard Opus rates, while standard interactions follow the 3.5 Sonnet pricing structure.

Technical requirements for implementation

Accessing the model requires a specific configuration for data retention, which can be performed using a simple cURL request or the AWS CLI. By executing a PUT request to the Bedrock data retention endpoint with the provider_data_share parameter, developers enable the necessary data sharing for the model to function. Once configured, developers can call the model using the Anthropic Messages API or the Bedrock Converse API. The model is currently deployed in the US East (N. Virginia) and Europe (Stockholm) AWS Regions, with broader regional support expected to follow in upcoming updates.

Key Technical Takeaways

• Model Availability: Currently accessible via Amazon Bedrock and the Claude Platform on AWS.
• Data Retention: Requires 30-day retention of inputs and outputs for misuse monitoring.
• Safety Routing: High-risk prompts in sensitive categories are routed to Claude 3 Opus.
• Vision Capability: Enhanced ability to parse and analyze visual data from tables, charts, and diagrams.

As of June 2026, the deployment of Claude 3.5 Sonnet marks a shift toward more autonomous, task-oriented AI agents within the AWS ecosystem. Organizations looking to adopt the technology should consult the Amazon Bedrock documentation for updated API reference materials and regional availability logs.