Claude Opus 4.5 Now on Amazon Bedrock

Anthropic's New Foundation model, Claude Opus 4.5, is now available on Amazon Bedrock, a fully managed service that offers a selection of top foundation models from leading AI companies. Opus 4.5 is a logical step forward in what AI systems can do and standardizes all codes, agents, computing, and office functions. It releases both sonnet 4,5 and opus 4.1 while providing Opus-level at One-One-Third cost.

In this post, I'll show you what makes this model different, go through key business applications, and show you how to use the new OPUS 4.5 tools from Amazon Bedrock. Finally, you will understand how to use the capabilities of the production agent routing model.

Claude Opus 4.5: What makes this model different

Opus 4.5 is ANTHROPIC's Advanced Model offering in the Opus class, designed for developers building AIs that are amazing AIs that can reason, plan, and execute complex tasks with minimal attention. It improves Sonnet 4.5 with better performance for existing use cases and adds new capabilities for complex workflows.

The model passes through the professional software business, reaching 80,9% on the SWECH bench confirmed, helping to turn multi-day development projects into hours of work. It works independently including multi-language development capabilities, as well as advanced behaviors such as code efficiency, better test coverage, and cleaner architecture decisions. Office production, the model handles complex projects in the end. It Powers Agents who create PowerPoint presentations, Excel spreadsheets, and Word documents with Professional Polish, including document restructuring contracts and NDAS. The model also produces high quality responsive and creative HTML. It maintains consistency and accuracy – important in financial and other industries where precision matters – and maintains context across files across long projects.

This is the best anthropic observation model at the moment, it reaches 80.7% in MMMU, so that it depends on the complex translation and analysis such as the analysis of complex tasks, or computer-based tasks – with the performance of computer operations

The model introduces two key improvements for Agent developers. The tools search tool allows agents to work with hundreds of tools by finding dynamic and loading only what you need instead of tens of thousands of tokens and block schemas exactly when scaling large libraries. Tool Using Examples allows you to provide sample calls that correspond directly to the tool description, improving the accuracy of complex schemas with organized objects or arrays.

Business applications and use cases

OPUS 4.5 Excels in the following use cases:

• Software development: Create agents that write and write code in all projects, manage the construction of full stacks, or agent agents that break high-level goals with high steps. This generation of Claude has spent a full development life: Opus 4.5 with production code and residential software agents, Sonnet 4.5 with sublioner e Scarment. OPUS 4.5 can analyze technical documents, plan the use of software, write the necessary code, and reverse with IT inferatives – while the context of the state in the whole process.
• Business activities and office activities: Manage complex projects from start to finish. Opus 4.5 uses memory to maintain context and file consistency, as well as improvements in creating spreadsheets, slides, and documents. The model manages ongoing, automated business projects.
• Financial Analysis: Work on all the complex information systems-control awards, market reports, internal data that enable forecasting, forecasting and effective compliance. The model's consistency and accuracy make it useful in finance and other industries where accuracy is an issue.
• The rise of cybersestiture: Deliver technical grade analysis of security operations performance, log coordination, security information databases, and security intelligence for security event detection and automated incident response.

Integration with AgentCore for Amazon Bedrock

Amazon Bedrock offers Enterprise Foundation for deploying Opus 4.5 in production. The fully managed service provides an integrated API for base models with Enterprise-grade security, compliance and management.

Opus 4.5 integrates with Amazon Bedrock's AgentCore, which provides the infrastructure and framework for building production agents. AgentCore features persistent memory to store context across sessions, Table Gateway for converting your APIs and lambda functions to compatible tools, and built-in identity management for secure access to resources. You can deploy and monitor complete agents with a full session session, long-term performance support (up to 8 hours), and monitoring features), so you can focus on building agents instead of managing infrastructure.

Amazon Bedrock Agentcore provides additional capabilities for production publishing. My Tool transforms your existing APIs and Lambda functions into compatible tools that correspond to a small code that works with fool's tool's tool tool search feature. Built-in monitoring using Amazon CloudWatch tracks Token Usage, latency, and error rates across your Agent Workflow.

Getting started

Access the Opus 4.5 model today through Amazon Bedrock. I will demonstrate the search capabilities of the model tool – a feature that allows agents to work with hundreds of tools without having to load all the definitions into context up front. First, I import the required modules and set up the Amazon Bedrock client:

# Import required libraries
import boto3
import json
# Create a session and Bedrock client
session = boto3.Session()
bedrock_client = session.client(
    service_name="bedrock-runtime",
    region_name="us-east-1"

For the sake of this example, I will describe several tools with defer_loading enabling search for tools. This allows the model to find and load only the tools it needs instead of loading all the definitions upfront:

# Define tools with tool search enabled
tools = [
    # Enable tool search - allows dynamic tool discovery
    {
        "type": "tool_search_tool_regex",
        "name": "tool_search_tool_regex"
    },
    # Tools marked with defer_loading are discovered on-demand
    {
        "name": "get_weather",
        "description": "Get current weather for a location",
        "input_schema": {
            "type": "object",
            "properties": {
                "location": {"type": "string"},
                "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]}
            },
            "required": ["location"]
        },
        "defer_loading": True,
        # Provide example inputs to improve accuracy for complex schemas
        "input_examples": [
            {"location": "San Francisco, CA", "unit": "fahrenheit"},
            {"location": "Tokyo, Japan", "unit": "celsius"}
        ]
    },
    {
        "name": "search_documentation",
        "description": "Search AWS documentation",
        "input_schema": {
            "type": "object",
            "properties": {
                "query": {"type": "string"},
                "service": {"type": "string"}
            },
            "required": ["query"]
        },
        "defer_loading": True,
        "input_examples": [
            {"query": "Lambda pricing", "service": "lambda"},
            {"query": "S3 bucket policies"}
        ]
    },
    {
        "name": "analyze_logs",
        "description": "Analyze application logs for errors",
        "input_schema": {
            "type": "object",
            "properties": {
                "log_file": {"type": "string"},
                "time_range": {"type": "string"}
            },
            "required": ["log_file"]
        },
        "defer_loading": True,
        "input_examples": [
            {"log_file": "/var/log/app.log", "time_range": "last 24 hours"},
            {"log_file": "/var/log/error.log"}
        ]
    }
]

Now I call the model using invoke_model API with the effort parameter set to it medium:

# Construct the request with beta features enabled
request_body = {
    "anthropic_version": "bedrock-2023-05-31",
    # Enable beta features: tool search, tool examples, and effort parameter
    "anthropic_beta": ["tool-search-tool-2025-10-19", "tool-examples-2025-10-29", "effort-2025-11-24"],
    "max_tokens": 4096,
    "temperature": 0.7,
    # Set effort to "medium" for balanced token usage
    "output_config": {
        "effort": "medium"
    },
    "messages": [
        {
            "role": "user",
            "content": "What's the weather in Seattle?"
        }
    ],
    "tools": tools
}

)
# Invoke the model
response = bedrock_client.invoke_model(
    modelId="global.anthropic.claude-opus-4-5-20251101-v1:0",
    body=json.dumps(request_body)

# Parse the response
response_body = json.loads(response['body'].read())

The model uses a tool search to find the right tool (get_weather) from the library without preloading all the tool definitions. The effort parameter, found in Beta, controls how the model spends toys with thinking beyond logic, tool calls, and responses. You can set the effort to high With excellent results, medium of moderate use, or low of the use of the saved token.

Key features of agent development

Opus 4.5 has several capabilities that make it ideal for production agents. The model preserves the consistency of decomposition across non-invariant decision flow tasks in which agents execute multi-step processes over hours or days. Better tool handling means agents interact reliably with external systems, APIs, and software interfaces – a model that selects the right tools and translates results more accurately. Opus 4.5 also tracks information throughout the conversation and preserves context, helping to gather information over time and make decisions based on history.

The effort parameter, available in Beta, gives you control over token usage. You can set high For good results where quality is high, medium with moderate performance, or low of the use of the saved token. Opus 4.5 transforms the use of tokens for all logic, tool calls, and responses based on this system. For production deployments, Amazon Bedrock AgentCore provides monitoring and recognition through CloudWatch integration, real-time parameter usage tracking (useful when latency metrics are detected, and cost-effective pricing.

To place a higher material

The model is priced at 5 percent by $ 5 by $ 25 by $ 25 by 60 by million tokens, which makes OPUS level intelligence available in a third of the previous offerings.

Availability and Access

This model is available today in Amazon Bedrock by using regional detection in Bedrock, which automatically affects requests for available power in AWS regions at peak times.

Use this model for agents that handle long-running tasks, connect multiple devices, or maintain context for extended sessions.

For detailed information on availability, pricing, and model specifications, visit the Amazon Bedrock documentation.

Lasting

This post showed you how to get started with Claude Opus 4.5 on Amazon Bedrock. OPUS 4.5 Excels in complex, time-consuming workflows such as software development and business operations. Opus 4.5's capabilities in tool handling, context management, and decision making make it extremely useful for construction agents working reliably in production environments. The model works well for agents in Software Engineering, change research, and business automation.

I encourage you to test Opus 4.5 for your agent workflow. Consider how their capabilities can improve manual processes in your organization, or support new forms of automation. The combination of the capabilities of Opus 4.5 with Amazon Bedrock's Enterprise Features Amazon provides a foundation for AI AI manufacturers.

To get started, try the model on the Amazon Bedrock Console, check out the technical documentation, and check out Anthropic's Claude Voice Model Cadel page for more about its capabilities. To deploy agents at scale, check out Opus 4.5 in Amazon Bedrock Agentcore for managed infrastructure with toolchestration and monitoring.

I'd love to hear what you build with this model – Share your experiences and use cases for agents in the comments below!

About the writers

Jonathan Evans Ingabe ukwakhiwa kwezixazululo emhlabeni wonke kwe-aw aw aw aw, lapho asiza khona amakhasimende athola ubuchwepheshe be-AI onqenqemeni ngamamodeli we-anththropic e-Anthropic e-Amazon Bedrock, ukuxazulula izinselelo zebhizinisi eziyinkimbinkimbi. With a background in AI / ML engineering and hands-on ownership of machine learning workflows in the cloud, Jonathan is passionate about making AI advanced and impactful for organizations of all sizes.