Introducing Claude Haiku 5.5 on AWS

As we speak, we’re excited to announce the supply of Claude Haiku 5.5 on Amazon Bedrock and Claude Platform on AWS. Based on Anthropic, Claude Haiku 5.5 is the quickest and best mannequin within the Claude 5.5 household, constructed for subagents and high-volume, cost-sensitive work. It additionally prices round 75 % lower than Claude Haiku 4.5 for many duties.

Amazon Bedrock offers you Haiku 5.5 capabilities whereas holding your knowledge inside AWS infrastructure with Regional knowledge residency. It really works with the AWS controls your crew already makes use of, together with AWS Id and Entry Administration (IAM) for entry, AWS CloudTrail for audit, Amazon CloudWatch for monitoring, and Amazon Bedrock Guardrails. Utilization seems in your AWS invoice.

Claude Platform on AWS offers you direct entry to Anthropic’s native platform expertise and capabilities by the AWS Administration Console. Construct, take a look at, and deploy with the identical APIs, options, and console expertise you’d get working with Anthropic straight, unified with AWS billing and authentication.

This submit covers Claude Haiku 5.5’s enhancements, sensible steering on when to decide on Haiku 5.5, and tips on how to get began on Amazon Bedrock.

What makes Claude Haiku 5.5 totally different

Claude Haiku 5.5 is Anthropic’s most succesful Haiku mannequin, throughout coding, device use, laptop use, and agentic duties. It’s additionally the primary Haiku mannequin with effort controls, so you’ll be able to tune price towards intelligence for every job as an alternative of choosing one setting for a whole workload.

The enhancements stand out on fast and repeatable work at scale. For coding duties, it acts as a subagent routing requests, reviewing code and classifying lengthy paperwork. For information work, Haiku 5.5 pulls key info from small-to-medium paperwork, does preliminary scans, and solutions fast questions over a information base. For interactive functions, it responds quick sufficient for easy conversations to have fast and useful solutions. It additionally handles conventional pure language processing (NLP) duties resembling classification, summarization, and textual content technology on the quantity and price that manufacturing options require.

Haiku 5.5 handles agentic coding and multi-step device use, and helps high-resolution pictures. Haiku 5.5 can be utilized as a powerful laptop use subagent for repetitive browser and desktop duties, at a value that holds up at scale. In improvement workflows, it’s a great match for iterating rapidly on UI and UX modifications and for small, particular code base modifications throughout a number of recordsdata.

Pairing Claude Haiku 5.5 with Opus 5.5

Haiku 5.5 pairs with the not too long ago introduced Claude Opus 5.5. Collectively, they make a powerful crew: Opus 5.5 plans and makes the judgment calls, and Haiku 5.5 carries out well-defined duties rapidly and at scale. You get cautious reasoning the place it counts and decrease price and latency all over the place.

Claude Opus 5.5 plans the work and makes the judgment calls. It breaks down advanced issues, decides the method, and takes on the toughest reasoning, resembling launch debugging, safety overview of enormous pull requests, and lengthy analyses that finish in a completed report.
Claude Haiku 5.5 takes on the quick layer of subagents. It handles fast, high-volume duties, resembling routing requests, classifying and summarizing, rewriting lengthy paperwork, and making use of small, particular modifications throughout many recordsdata. As a overview subagent, it may well rapidly examine the order of operations and high-level route, so Opus 5.5 can spend its tokens on the toughest reasoning. As a result of it’s quick and price environment friendly, you’ll be able to run many Haiku subagents in parallel.

Getting began with Claude Haiku 5.5 on Amazon Bedrock

To strive Haiku 5.5, open the Amazon Bedrock console, select Take a look at, then Playground, and choose Haiku 5.5 because the mannequin. From there, you’ll be able to run a immediate straight towards it.

Determine 1: Deciding on an Anthropic Claude mannequin within the Amazon Bedrock console Playground

Programmatically, you’ll be able to name the mannequin with the Anthropic Messages API towards bedrock-runtime by the Anthropic SDK. You too can use the Invoke and Converse APIs on bedrock-runtime by the AWS Command Line Interface (AWS CLI) and AWS SDK.

Stipulations

It’s essential to have the next conditions:

Lively AWS account with Amazon Bedrock entry.
AWS Command Line Interface (AWS CLI) put in and configured.
Python 3.10+.
Boto3 put in: pip set up boto3.
Anthropic SDK put in: pip set up anthropic.
The Amazon Bedrock Token Generator for Amazon Bedrock authentication put in: pip set up aws_bedrock_token_generator.
AWS Id and Entry Administration (IAM) permissions: bedrock:InvokeModel, bedrock:InvokeModelWithResponseStream.

Right here’s a fast instance utilizing the AWS SDK for Python (Boto3) with the InvokeModel API:

import boto3
import json

# Create a Bedrock Runtime shopper
bedrock_runtime = boto3.shopper(
    service_name="bedrock-runtime",
    region_name="us-east-1"
)

# Invoke Claude Haiku 5.5
response = bedrock_runtime.invoke_model(
    modelId="international.anthropic.claude-haiku-5-5",
    contentType="software/json",
    settle for="software/json",
    physique=json.dumps({
        "anthropic_version": "bedrock-2023-05-31",
        "max_tokens": 4096,
        "messages": [
            {
                "role": "user",
                "content": "Can you explain the features of Amazon Bedrock?"
            }
        ]
    })
)

consequence = json.masses(response["body"].learn())
# Haiku 5.5 could return a considering block earlier than the textual content block,
# so choose the textual content block quite than a hard and fast index.
print(subsequent(b["text"] for b in consequence["content"] if b["type"] == "textual content"))

You too can use the Amazon Bedrock Converse API for a unified multi-model expertise:

import boto3

# Create a Bedrock Runtime shopper
bedrock_runtime = boto3.shopper(
    service_name="bedrock-runtime",
    region_name="us-east-1"
)

# Invoke Claude Haiku 5.5
response = bedrock_runtime.converse(
    modelId="international.anthropic.claude-haiku-5-5",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "text": "Can you explain the features of Amazon Bedrock?"
                }
            ]
        }
    ],
    inferenceConfig={
        "maxTokens": 4096
    }
)

if 'output' in response:
    blocks = response['output']['message']['content']
    print('n'.be a part of(b.get('textual content', '') for b in blocks if 'textual content' in b))

You too can use the Anthropic Messages API by the anthropic SDK bundle for a streamlined expertise:

from anthropic import Anthropic
from aws_bedrock_token_generator import provide_token

token = provide_token(area="us-east-1")

shopper = Anthropic(
    base_url="https://bedrock-runtime.us-east-1.amazonaws.com/anthropic",
    api_key=token,
)

# Invoke Claude Haiku 5.5
response = shopper.messages.create(
    mannequin="international.anthropic.claude-haiku-5-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Can you explain the features of Amazon Bedrock?"}],
)
print(response)

You may discover the Getting Began pocket book for extra examples. You may monitor utilization, efficiency, and prices by Amazon CloudWatch and AWS Price Explorer to scale your functions as demand grows.

Availability

Claude Haiku 5.5 is out there right this moment on Amazon Bedrock by the US Geo CRIS (us.), EU Geo CRIS (eu.), AU Geo CRIS (au.), JP Geo CRIS (jp.) and World CRIS (international.) inference profiles on bedrock-runtime. In AWS GovCloud (US), it’s out there on each the bedrock-runtime and bedrock-mantle endpoints.

See the Amazon Bedrock documentation for the total checklist of supported AWS Areas. For pricing info, see Amazon Bedrock pricing. It’s additionally out there by Claude Platform on AWS in North America

Give Claude Haiku 5.5 a strive on the Amazon Bedrock console, in Claude Platform on AWS, or discover the Getting Began notebooks on GitHub.