Claude Opus 4.6: Anthropic’s powerful model for coding, agents, and enterprise workflows is now available in Microsoft Foundry

With the introduction of Claude Opus 4.6 in Microsoft Foundry, developers can now assign intricate tasks from start to finish, relying on AI to carry out operations independently in a production environment.

At Microsoft, we are convinced that intelligence and trust are essential for deploying agentic AI at scale. Powered by Azure, Microsoft Foundry combines these elements on a robust, secure cloud platform for enterprise AI. With the launch of Claude Opus 4.6, we enhance our agents’ ability to learn from and interact with business systems more effectively.

Claude Opus 4.6 incorporates Anthropic’s most sophisticated reasoning capabilities into Microsoft Foundry, creating a platform where intelligence meets trust to facilitate autonomous operations. Within Foundry, Opus 4.6 taps into various knowledge sources through Foundry IQ, utilizing data from M365 Work IQ, Fabric IQ, and the web. This model excels in complex tasks across coding, knowledge work, and agent-guided workflows, delivering enhanced reasoning skills and dependable instruction adherence.

Now, developers can automate their most complex assignments, managing everything from requirements gathering to implementation and maintenance. Business users can create polished documents, conduct thorough research, and write content with a professional edge and industry insight.

At Adobe, we are continually assessing new AI capabilities that can help us provide powerful and intuitive experiences for our users. We’ve been testing Claude models in Microsoft Foundry and are thrilled about the progress of Anthropic’s model roadmap. Foundry offers us a versatile, enterprise-ready space to explore advanced models while ensuring the trust and governance that are so critical for Adobe.

—Michael Marth, VP Engineering for Experience Manager and LLM Optimizer, Adobe

Presenting Opus 4.6: Frontier intelligence for real-world applications

Claude Opus 4.6 represents the latest iteration of Anthropic’s most advanced model and is considered the optimal choice for coding, enterprise agents, and professional tasks. Featuring a 1M token context window (beta) and a max output of 128K tokens, Opus 4.6 is perfectly suited for:

• Production coding
• Advanced agents
• Office duties
• Financial evaluations
• Cybersecurity
• General computer tasks

By merging Anthropic’s leading model with Foundry’s comprehensive tools, clients can transition from testing to production seamlessly, without the hassle of piecemeal infrastructure.

1. Elevating autonomous coding

Opus 4.6 manages extensive codebases adeptly and excels at prolonged tasks such as refactoring, bug detection, and intricate implementations.

This enables seasoned engineers to offload tasks that typically take days, allowing them to concentrate on reviews and strategic decision-making instead of getting bogged down in specifics.

Utilising Foundry’s managed setup and operational controls, teams can significantly reduce development timelines from days to mere hours, while upholding the necessary rigor for real-world production systems.

Macroscope has relied on Anthropic’s cutting-edge models to transform AI code reviews, helping us identify critical bugs before they reach production. We’re excited to enhance this partnership by scaling on Azure infrastructure and providing our software to teams globally.

—Kayvon Beykpour, CEO & Co-Founder, Macroscope

2. Enhanced knowledge work across the organisation

Opus 4.6 marks a significant improvement in how enterprises manage knowledge tasks. Anthropic identifies this process through three key areas: searching, analyzing, and creating. Within Microsoft Foundry, these functionalities can be directly applied to enterprise data, workflows, and agent-driven applications.

Opus 4.6 grasps the standards and conventions of various professional fields, generating documents, spreadsheets, and presentations that reflect the quality of expert work. Coupled with Foundry’s governance and accessibility features, Opus 4.6 is highly suitable for finance, legal, and other critical industries where quality, context, and adherence to guidelines are paramount.

At Dentons, we are implementing generative AI across drafting, reviewing, and research workflows within our global teams. Claude in Microsoft Foundry provides the advanced reasoning we require for legal tasks, underpinned by the governance and operational protocols essential in a corporate setting. Improved model reasoning reduces unnecessary revisions and enhances consistency, allowing our lawyers to focus on higher-value judgments.

—Matej Jambrich, CTO, Dentons Europe

In-depth financial analysis

Opus 4.6 shines in connecting insights from regulatory documents, market reports, and internal enterprise data, revealing conclusions that would typically take analysts days to piece together.

Its advanced reasoning capabilities empower it to:

• Navigate intricate financial and regulatory environments
• Produce compliance-sensitive content
• Ensure consistency and traceability across complex analytical procedures

When utilized through Microsoft Foundry, these processes benefit from Azure’s security, compliance, and auditing capabilities, helping organisations confidently implement frontier AI in high-stakes analyses.

3. Progressing agentic computer use abilities

Anthropic asserts that Opus 4.6 delivers remarkable advancements in computer utilisation, performing strongly on industry benchmarks for visual comprehension and multi-step navigation. Opus 4.6 establishes a new benchmark for computer operations. Claude can now operate computers more accurately, tackle complex tasks, and smoothly work across various applications.

It can interact with software, navigate user interfaces, fill in forms, and transfer data between applications. When implemented within Microsoft Foundry, these processes can function as secure, governed agents, facilitating the automation of workflows that encompass legacy systems, document handling, and operational tools.

Opus 4.6 can oversee intricate, multi-tool workflows with minimal supervision—an essential need for teams managing AI systems at scale.

Anthropic on Azure provides Momentic with the reliability we need to process millions of tokens every hour with state-of-the-art models like Opus 4.5. Azure’s platform integrates seamlessly with Anthropic’s SDK, even supporting beta features like reasoning out of the box.

—Jeff Ann, CTO & Co-Founder, Momentic AI

4. Security and high-stakes reasoning for agents

Opus 4.6 is highly adept at handling agentic workflows, orchestrating intricate tasks across multiple tools reliably. It can autonomously create sub-agents, distribute tasks simultaneously, and push projects forward with minimal oversight.

In security tasks, Opus 4.6 provides deep reasoning abilities, helping teams detect subtle patterns and complex attack points with high precision.

Anthropic is a trusted ally for governments and companies alike. Their speed and accuracy are already enabling Everstar to make swift, secure nuclear energy deployments a reality. I’m eager to see these features integrated directly on Azure for safe implementations for our government and nuclear clientele.

—Kevin Kong, Founder & CEO, Everstar

New API features launching with Opus 4.6

Alongside Opus 4.6, Anthropic is unveiling new API features via Microsoft Foundry that provide developers with improved control, scalability, and efficiency:

• Adaptive reasoning: This allows Claude to intelligently decide when and how much reasoning is necessary, enhancing performance on simpler tasks while dedicating more thought to complex challenges.
• Context Compaction (beta): This feature supports ongoing conversations and agentic workflows by summarising older context when token limits are reached.
• 1M Context Window (beta): Offers support for 1M tokens, with premium fees applicable for usage above 200K tokens.
• Max effort control: A new max effort setting enhances granularity in token allocation across reasoning, tools, and output.
• 128K Output Tokens: This allows for richer, more detailed responses in a single output.

Claude Opus 4.6 in Microsoft Foundry

As AI systems evolve from support roles to autonomous functions, achieving success relies on more than just advanced intelligence; it necessitates trustable intelligence to operate effectively in real-world settings. Claude Opus 4.6 brings sophisticated reasoning and extended execution capabilities, while Microsoft Foundry provides the contextual infrastructure for responsibly applying that intelligence at scale.

Through Claude in Foundry, organisations can transcend isolated trials and steer towards intelligent, agent-driven systems that yield tangible business benefits—anchored in trust, governance, and operational discipline.

Opus 4.6 is also accessible through Microsoft Copilot Studio, enabling companies to swiftly build, manage, and deploy advanced AI agents without needing custom coding.