Claude Sonnet 5 Now Available: Stronger, More Stable AI for Programming and Agents

Claude Sonnet 5, the latest model in Anthropic's Sonnet series, is now generally available on Amazon Bedrock (overseas regions) and the Claude Platform, delivering a substantial capability jump while retaining Sonnet‑class pricing.

The model approaches Opus‑level intelligence and continues the Sonnet series' balance of ability, cost, and speed, making it suitable for large‑scale programming, agent orchestration, and professional tasks. When cost‑sensitive, developers can choose Sonnet 5; for the highest performance where the price premium is justified, Opus remains the option.

Key differentiators include:

Programming: Sonnet 5 can navigate real codebases, perform multi‑file modifications, and carry out lengthy debugging and refactoring with less supervision, producing cleaner, more maintainable code.

Autonomous agents: It reliably handles complex dependency chains and multi‑step tool calls, fitting both customer‑facing and internal agents.

Professional work: The model transforms lengthy, unstructured information into briefings, analysis reports, and other structured deliverables, marking a clear upgrade over Sonnet 4.6.

Industry scenarios highlighted include financial services (spreadsheet modeling, financial analysis, self‑checking report agents), office productivity (high‑consistency report writing, document drafting, structured analysis), and automation of browser or desktop workflows. In production‑grade agent pipelines, Sonnet 5 can serve as the core inference engine, supporting tool invocation and unattended multi‑step tasks.

Quick‑start guide: Prerequisites are an AWS account with Bedrock access, the AWS CLI configured, Python 3.8+, boto3, and the Anthropic SDK ( pip install anthropic[bedrock]). Example code shows how to invoke the model via the Bedrock Runtime invoke_model API and the converse API, including request payloads, token limits, and response handling.

import boto3
import json

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

# Call Claude Sonnet 5
response = bedrock_runtime.invoke_model(
    modelId="us.anthropic.claude-sonnet-5",
    contentType="application/json",
    accept="application/json",
    body=json.dumps({
        "anthropic_version": "bedrock-2023-05-31",
        "max_tokens": 4096,
        "messages": [
            {
                "role": "user",
                "content": "Design a distributed architecture on AWS in Python that should support 100k requests per second across multiple geographic regions."
            }
        ]
    })
)
result = json.loads(response["body"].read())
print(result["content"][0]["text"])

The same request can be made with the Converse API:

response = bedrock_runtime.converse(
    modelId="us.anthropic.claude-sonnet-5",
    messages=[
        {
            "role": "user",
            "content": [{"text": "Design a distributed architecture on AWS in Python that should support 100k requests per second across multiple geographic regions."}]
        }
    ],
    inferenceConfig={"maxTokens": 4096}
)
print(response["output"]["message"]["content"][0]["text"])

For a more concise experience, the Anthropic anthropic[bedrock] SDK can be used:

from anthropic import AnthropicBedrockMantle

mantle_client = AnthropicBedrockMantle(aws_region="us-east-1")
message = mantle_client.messages.create(
    model="anthropic.claude-sonnet-5",
    max_tokens=4096,
    messages=[{"role": "user", "content": "Design a distributed architecture on AWS in Python that should support 100k requests per second across multiple geographic regions"}]
)
print(message.content[0].text)

Claude Sonnet 5 is now live across Amazon Bedrock’s supported regions and the Claude Platform in North America, South America, Europe, and APAC, enabling developers to experiment directly in the Bedrock console or via the SDKs.