Claude Haiku 5.5 arrives on Amazon Bedrock and Claude Platform on AWS

What Haiku 5.5 offers

AWS says Claude Haiku 5.5 costs around 75 percent less than Claude Haiku 4.5 for most tasks. It describes the model as Anthropic's most capable Haiku model across coding, tool use, computer use and agentic tasks.

It is also the first Haiku model with effort controls. These let developers tune cost against intelligence for each task, rather than choosing one setting for an entire workload.

The post says the gains show most on quick, repeatable work at scale. In coding, the model can act as a subagent that routes requests, reviews code and classifies long documents. For knowledge work, it can pull key information from small-to-medium documents, run initial scans and answer quick questions over a knowledge base. It also handles traditional natural language processing tasks such as classification, summarization and text generation, and it supports high-resolution images.

AWS also points to computer use, where Haiku 5.5 can serve as a subagent for repetitive browser and desktop tasks. In development workflows, it suits quick iteration on UI and UX changes and small, specific code base changes across multiple files.

Pairing with Opus 5.5

AWS recommends using Haiku 5.5 together with the recently announced Claude Opus 5.5. In this setup, Opus 5.5 plans the work and makes the judgment calls. It breaks down complex problems, decides the approach and handles the hardest reasoning, such as release debugging, security review of large pull requests and long analyses that end in a finished report.

Haiku 5.5 takes the fast layer of subagents: routing requests, classifying and summarizing, rewriting long documents and applying small changes across many files. As a review subagent, it can quickly check the order of operations and high-level direction, so Opus 5.5 can spend its tokens on the hardest reasoning. Because Haiku is fast and cost efficient, many subagents can run in parallel.

Getting started and data controls

On Amazon Bedrock, AWS says data stays within AWS infrastructure with Regional data residency. The model works with controls teams may already use: AWS Identity and Access Management for access, AWS CloudTrail for audit, Amazon CloudWatch for monitoring and Amazon Bedrock Guardrails. Usage appears on the AWS bill.

To try it, users can open the Amazon Bedrock console, choose Test, then Playground, and select Haiku 5.5. Programmatic access is available through the Anthropic Messages API against bedrock-runtime via the Anthropic SDK, and through the Invoke and Converse APIs using the AWS CLI and AWS SDK. The post includes Python examples using the model ID global.anthropic.claude-haiku-5-5. It notes that the model may return a thinking block before the text block, so code should select the text block instead of relying on a fixed index.

Prerequisites include an active AWS account with Amazon Bedrock access, the AWS CLI, Python 3.10+, Boto3, the Anthropic SDK and the Amazon Bedrock Token Generator. Required IAM permissions are bedrock:InvokeModel and bedrock:InvokeModelWithResponseStream.

Availability and what it means

Claude Haiku 5.5 is available on Amazon Bedrock through the US, EU, AU, JP and Global CRIS inference profiles on bedrock-runtime. In AWS GovCloud (US), it is offered on both the bedrock-runtime and bedrock-mantle endpoints. The post refers readers to the Amazon Bedrock documentation for supported Regions and to the pricing page for costs. It is also available through Claude Platform on AWS in North America.

Claude Platform on AWS gives direct access to Anthropic's native platform experience through the AWS Management Console, with the same APIs, features and console experience, unified with AWS billing and authentication.

For teams already building on AWS, the release offers a lower-cost model for high-volume work and a defined pattern for combining it with Opus 5.5. Actual savings and performance will depend on the workload, and AWS directs users to its pricing page for details.