Claude Sonnet 4.6 is now available on Amazon Bedrock

On February 17, 2026, Anthropic's latest model Claude Sonnet 4.6 became available on Amazon Bedrock. This article introduces its main features, how to use it on Bedrock, and verification results for the new "Context Compaction" feature.

https://www.aboutamazon.com/news/aws/anthropic-claude-4-opus-sonnet-amazon-bedrock?

Main Features of Claude Sonnet 4.6

• Anthropic's best computer use model: Achieved the highest score of any Sonnet model on the OSWorld benchmark
• Significantly improved coding performance: 79.6% on SWE-bench Verified. In early testing with Claude Code, it was preferred 70% of the time over Sonnet 4.5 and 59% over Opus 4.5
• 1M token context window (beta)
• Support for adaptive thinking / extended thinking
• Significantly improved prompt injection resistance

For detailed benchmark results, refer to the Anthropic official announcement.

How to Use on Amazon Bedrock

Inference Profile and Model ID

Sonnet 4.6 is accessed via an Inference Profile. Instead of specifying the base model ID directly, you specify the Cross-Region Inference profile ID as the modelId. If your existing code uses the base model ID, you'll need to replace it.

Also, starting with Sonnet 4.6, the naming convention has changed, with date and version numbers omitted.

# Previous (Sonnet 4.5)
anthropic.claude-sonnet-4-5-20250929-v1:0

# Sonnet 4.6
anthropic.claude-sonnet-4-6

You can check available profiles with the following command:

aws bedrock list-inference-profiles --region us-west-2

It was supported in the Tokyo region from the first day of release. The JP profile routes to Tokyo and Osaka, making it easier to implement in projects with domestic region restrictions.

IAM Permissions

When calling via an Inference Profile, you specify the profile's ARN in the Resource of the IAM policy. This differs from the traditional base model ARN, so existing policies should be checked.

{
    "Effect": "Allow",
    "Action": "bedrock:InvokeModel",
    "Resource": "arn:aws:bedrock:ap-northeast-1:*:inference-profile/jp.anthropic.claude-sonnet-4-6"
}

Python SDK Sample Code

I confirmed operation with the JP profile in the Tokyo region.

import boto3
import json

bedrock = boto3.client('bedrock-runtime', region_name='ap-northeast-1')

response = bedrock.invoke_model(
    modelId='jp.anthropic.claude-sonnet-4-6',
    body=json.dumps({
        "anthropic_version": "bedrock-2023-05-31",
        "max_tokens": 1024,
        "messages": [
            {"role": "user", "content": "こんにちは！あなたのモデル名とバージョンを教えてください。"}
        ]
    })
)

body = json.loads(response['body'].read())
print(body['content'][0]['text'])

Testing "Context Compaction" (Beta) on Bedrock

"Context Compaction" is a feature that automatically summarizes long conversation contexts, effectively extending the context length. While Anthropic's API documentation only lists Opus 4.6 as a supported model, the Sonnet 4.6 announcement explicitly states it's supported. I verified whether it actually works on Bedrock.

Verification Code

I sent about 58,000 tokens of dummy text and checked if compaction would trigger at a threshold of 50,000 tokens.

import boto3
import json

bedrock = boto3.client('bedrock-runtime', region_name='us-west-2')

base_text = (
    'Amazon Bedrock is a fully managed service that offers a choice of '
    'high-performing foundation models from leading AI companies like '
    'Anthropic Meta Mistral AI and Amazon through a single API. '
    'It provides serverless experience so you can get started quickly '
    'privately customize foundation models with your own data and easily '
    'integrate and deploy them into your applications using AWS tools '
    'and capabilities. '
)

messages = [
    {"role": "user", "content": base_text * 800},
    {"role": "assistant", "content": "I see a large amount of text about Amazon Bedrock."},
    {"role": "user", "content": "Summarize our conversation."}
]

response = bedrock.invoke_model(
    modelId='us.anthropic.claude-sonnet-4-6',
    body=json.dumps({
        "anthropic_version": "bedrock-2023-05-31",
        "anthropic_beta": ["compact-2026-01-12"],
        "max_tokens": 4096,
        "messages": messages,
        "context_management": {
            "edits": [{
                "type": "compact_20260112",
                "trigger": {"type": "input_tokens", "value": 50000}
            }]
        }
    })
)

body = json.loads(response['body'].read())

for block in body['content']:
    print(f"type: {block['type']}")
    if block['type'] == 'compaction':
        print(f"summary: {block['content'][:200]}...")
    elif block['type'] == 'text':
        print(f"text: {block['text'][:200]}...")

print(f"\nusage.iterations: {json.dumps(body['usage']['iterations'], indent=2)}")

Verification Results

I confirmed that compaction triggered with Sonnet 4.6. A type: compaction block was returned at the beginning of the response's content array, with an automatically generated summary of the conversation.

usage.iterations record:

1. compaction: Input 58,601 tokens → Generated 389 token summary
2. message: Generated final response from 479 tokens of context after summarization

58,601 tokens were compressed down to 479 tokens, working as intended. Since summary generation for compaction incurs additional token consumption, the entire usage.iterations should be aggregated for cost calculation.

Pricing

Same price as Sonnet 4.5 (input $3.00 / output $15.00 per 1M tokens), with significantly improved performance at the same price point.

Cost optimization tips:

• Adjust cost and quality balance with the effort setting in adaptive thinking
• Get 50% discount from on-demand pricing with batch inference

Summary

Claude Sonnet 4.6 is available in the Tokyo region from day one. You can immediately benefit from the new model by simply replacing the model ID in your existing codebase with the Inference Profile (jp.anthropic.claude-sonnet-4-6).

While maintaining the same price as the previous model, it offers dramatically improved coding performance and agent capabilities, and I've confirmed the implementation of the compaction feature. With such performance improvements at the same cost, I recommend actively testing and migrating systems currently using Sonnet 4.5.

Reference Links

• AWS Bedrock Anthropic page
• About Amazon announcement
• Anthropic official announcement - Introducing Sonnet 4.6
• Anthropic Context Compaction documentation