Introducing Claude Opus 4.5 in Microsoft Foundry

We’re thrilled to introduce Anthropic’s latest innovation, Claude Opus 4.5, now live in Microsoft Foundry! You can now access Opus 4.5 through the Microsoft Foundry, GitHub Copilot paid plans, and Microsoft Copilot Studio.

We find ourselves at a pivotal moment in the AI realm, transitioning from simple tools to true partners in our work. These models can grasp objectives, consider their limitations, and perform intricate workflows using multiple tools. They not only assist processes but also help to refine them for enhanced reliability, scalability, and efficiency.

Anthropic’s new model, Claude Opus 4.5, is a representation of this evolution. We’re excited to confirm that Opus 4.5 is now in public preview at Microsoft Foundry, GitHub Copilot paid plans, and Microsoft Copilot Studio.

Building on our recent announcement about the enhanced partnership with Anthropic, Microsoft Foundry is focused on providing Azure users with the most extensive range of advanced AI models in the cloud. Foundry aims to boost innovation with an AI platform that is integrated, secure, and easy to scale for AI applications and agents.

We’re thrilled to utilise Anthropic Claude models from Microsoft Foundry. The combination of Claude’s advanced reasoning capabilities alongside GPT models in one platform allows us the flexibility to create scalable, enterprise-level workflows that surpass mere prototypes.

—Michele Catasta, President, Replit

Opus 4.5 for Real Work

Opus 4.5 redefines coding standards, workflows, and productivity in the enterprise space by outperforming both Sonnet 4.5 and Opus 4.1, all at a more accessible price. Its adaptability for software engineering, advanced reasoning, tool utilisation, and visual tasks opens the door for organisations to innovate systems, automate key processes, and achieve quicker returns on investment.

By quickly integrating the latest models, Foundry helps Azure users stay ahead and fully leverage their agentic AI systems, all while ensuring centralised governance, security, and scalability.

1. Designed for Production Engineering and Agentic Capabilities

According to Anthropic, Opus 4.5 shines in industry-standard software engineering benchmarks, achieving impressive results on SWE-bench (80.9%). Early users describe the model as adept at deciphering unclear requirements, evaluating architectural choices, and finding fixes for cross-system issues.

Claude Opus 4.5 benchmark results from Anthropic

Opus 4.5 stands out as one of the most effective tool-using models currently available, facilitating agents that efficiently operate across numerous tools. Developers can now leverage several significant enhancements:

• Programmatic Tool Calling: Execute tools directly in Python for more effective, predictable workflows.
• Tool Search: Effortlessly discover tools from extensive libraries without clogging up the context window.
• Tool Use Examples: Precise tool calling for sophisticated tool schemas.

Collectively, these features enable advanced agents in fields like cybersecurity, software engineering, and financial modelling, all of which require intricate tool interactions. Opus 4.5 demonstrates strong, real-world intelligence when applying these tools creatively within set constraints. In tests, the model effectively managed complex issues, such as airline change rules, optimising outcomes through careful handling of upgrades, downgrades, cancellations, and rebookings. This level of adaptive problem-solving marks a significant advance in what agentic AI can achieve.

Manus extensively uses Anthropic’s Claude models for their exceptional capabilities in coding and long-term task planning, in addition to handling agentic tasks. We’re immensely excited to employ them on Microsoft Foundry!

—Tao Zhang, Co-founder & Chief Product Officer, Manus AI

2. Enhanced Developer Experience on Foundry

Opus 4.5, combined with the new capabilities available on Foundry, is designed to support teams in creating more effective agentic systems:

• Effort Parameter (Beta): Manage the computational effort Claude assigns for thinking, tool calls, and responses to optimise performance, latency, and costs according to your specific requirements.
• Compaction Control: Improve the handling of long-running agentic tasks with new SDK tools for context management over extended activity.

These enhancements enable greater predictability and operational control for enterprise workloads.

3. Boosted Office Productivity and Computer Use

Opus 4.5 also excels as Anthropic’s leading vision model, enhancing workflows that rely on intricate visual analysis and multi-step navigation. The performance in computer usage has seen a notable boost, allowing for more reliable automation of desktop tasks.

For knowledge workers, the model significantly enhances the ability to generate spreadsheets, presentations, and documents. It delivers consistent, professional-grade outputs while demonstrating a genuine understanding of the domain, making it suitable for finance, legal, and other precision-critical sectors. The model effectively leverages memory to maintain context and reliability across files during complex projects.

4. Safety and Security

As stated by Anthropic, Opus 4.5 features substantial advancements in safety and security. The model has a lower rate of misaligned responses, exhibits stronger resilience against prompt-injection attacks, and shows more reliable performance in complex tasks.

These enhancements align with Microsoft’s dedication to providing enterprise clients with models that adhere to strict standards for safety, governance, and operational integrity.

Use Cases

Opus 4.5 caters to a variety of applications:

• Software Development: Deploy agents capable of managing complex, multi-system development tasks with minimal oversight.
• Financial Analysis: Integrate insights across regulatory filings, market reports, and internal data to create advanced predictive models and ensure proactive compliance monitoring.
• Cybersecurity: Combine logs, vulnerability databases, and threat intelligence for top-notch threat detection and automated incident responses.
• Enterprise Operations: Oversee intricate workflows requiring coordination between multiple tools, systems, and data sources.

Pricing and Availability

Opus 4.5 delivers cutting-edge performance, setting a new standard for various applications at just one-third of the price of earlier Opus-class models.

Get Started Today

You can now access Claude Opus 4.5 in Microsoft Foundry and soon in Visual Studio Code via the Foundry extension. Head over to the Foundry portal to start building with Opus 4.5 today!