According to an official AWS announcement, Claude Haiku 5.5 is now available on Amazon Bedrock and Claude Platform on AWS. According to Anthropic, Claude Haiku 5.5 is the fastest and most efficient model in the Claude 5.5 family, built specifically for subagents and high-volume, cost-sensitive work. In addition, the model costs around 75 percent less compared to Claude Haiku 4.5 for most tasks.
Amazon Bedrock provides Claude Haiku 5.5 capabilities while keeping data within AWS infrastructure with Regional data residency. The service operates with AWS controls, including AWS Identity and Access Management (IAM) for access management, AWS CloudTrail for audit purposes, Amazon CloudWatch for monitoring, and Amazon Bedrock Guardrails. Usage of the model appears directly on the AWS bill.
Concurrently, Claude Platform on AWS provides direct access to Anthropic’s native platform experience and capabilities through the AWS Management Console. The solution enables building, testing, and deploying applications using the same APIs, features, and console interface available when working directly with Anthropic, unified with AWS billing and authentication mechanisms.
What Makes Claude Haiku 5.5 Different
According to details in the AWS post, Claude Haiku 5.5 is Anthropic’s most capable Haiku model across coding, tool use, computer use, and agentic tasks. It is also the first Haiku model to include effort controls, allowing users to tune the ratio of cost against intelligence for each task individually, rather than having to rely on a single setting across an entire workload.
The model's improvements stand out in fast, repeatable work at scale. For coding tasks, the model acts as a subagent that routes requests, reviews code, and classifies long documents. In knowledge tasks, the model extracts key information from small-to-medium documents, performs initial scans, and answers quick questions over knowledge bases. For interactive applications, the model responds with speed that allows simple conversations to receive rapid and helpful answers. In addition, it handles traditional natural language processing (NLP) tasks such as classification, summarization, and text generation at the volume and cost required for production features.
Claude Haiku 5.5 handles agentic coding tasks and multi-step tool use, and supports high-resolution images. The model can serve as a computer use subagent for repetitive browser and desktop tasks at a cost that holds up at large scale. In development workflows, it is suited for rapid iterations on user interface and user experience (UI/UX) changes and for executing small, specific codebase changes across multiple files.
Pairing Claude Haiku 5.5 with Claude Opus 5.5
According to the publication, Claude Haiku 5.5 pairs with the recently announced Claude Opus 5.5 model. The pairing between them creates a division of roles where Opus 5.5 plans and makes judgment calls, while Haiku 5.5 carries out well-defined tasks quickly and at scale. This structure provides rigorous reasoning capability where needed, alongside lower costs and reduced latency in the remaining stages.
In this division of labor, Claude Opus 5.5 plans the work and makes the decisions, breaks down complex problems, decides the approach, and takes on the most difficult reasoning operations, such as release debugging, security reviews of large pull requests, and extended analyses resulting in a finished report.
On the other hand, Claude Haiku 5.5 takes on the fast subagent layer. It handles fast, high-volume tasks such as routing requests, classification and summarization, rewriting long documents, and applying small, specific changes across multiple files. As a review subagent, it can quickly check the order of operations and high-level direction, allowing Opus 5.5 to spend its tokens on the hardest reasoning. Due to its speed and cost efficiency, multiple Haiku subagents can be run in parallel.
Getting Started and Prerequisites on Amazon Bedrock
To try the model through the user interface, users can open the Amazon Bedrock console, choose Test, select Playground, and choose Haiku 5.5 as the active model to run a direct prompt. Programmatically, the model can be called using the Anthropic Messages API against the bedrock-runtime endpoint through the Anthropic SDK, or by using the Invoke API and Converse API against bedrock-runtime via the AWS CLI and AWS SDK.
The prerequisites for using the model include:
- An active AWS account with access to Amazon Bedrock.
- AWS Command Line Interface (AWS CLI) installed and configured.
- A Python environment version 3.10 or higher.
- Installation of the Boto3 library (via the command
pip install boto3). - Installation of the Anthropic SDK (via the command
pip install anthropic). - Installation of the Amazon Bedrock Token Generator for authentication (via the command
pip install aws_bedrock_token_generator). - Permissions in AWS Identity and Access Management (IAM):
bedrock:InvokeModelandbedrock:InvokeModelWithResponseStream.
Integration Examples and Regional Availability
The post presents Python code examples: invoking via InvokeModel passes the model identifier global.anthropic.claude-haiku-5-5 with a JSON-formatted request body and extracts the text block from the response (taking into account that Haiku 5.5 may return a thinking block before the text block). When using the Converse API, a unified multi-model structure is obtained with inference configuration (inferenceConfig). When using the Anthropic SDK, a token is generated via the token generator and the call is made to the base URL of bedrock-runtime in the us-east-1 region.
To track system activity, usage, performance, and costs can be monitored using Amazon CloudWatch and AWS Cost Explorer in order to scale applications as demand grows.
Claude Haiku 5.5 is available starting today on Amazon Bedrock through the US Geo CRIS (us.), EU Geo CRIS (eu.), AU Geo CRIS (au.), JP Geo CRIS (jp.), and Global CRIS (global.) inference profiles on bedrock-runtime. In AWS GovCloud (US), the model is available on both the bedrock-runtime and bedrock-mantle endpoints. In addition, the model is available through Claude Platform on AWS in the North America region. The full list of regions and pricing details are available in the official Amazon Bedrock documentation.