Anthropic announced today that it has expanded its strategic partnership with Microsoft. Its three major artificial intelligence models, Claude Sonnet 4.5, Haiku 4.5 and Opus 4.1, have been officially launched for public testing on Microsoft Foundry. Azure customers can now build enterprise-level production applications and intelligent agents through the Foundry platform, taking full advantage of the AI advantages in the Microsoft ecosystem.
This cooperation allows enterprises to directly deploy Claude models in the familiar Microsoft environment and apply them to many scenarios such as code development and smart office. Not only that, developers can also use Claude Code, an AI programming assistant in Microsoft Foundry to improve project development automation.
In addition to the original integration in Microsoft 365 Copilot - such as the Researcher multi-step research agent in the Copilot Studio platform, and custom AI agent functions - Microsoft Excel's Agent Mode now also supports calling Claude. Users can use Claude directly in Excel tables to generate formulas, analyze data, find errors and optimize solutions.
For enterprise users who have deeply invested in Microsoft Foundry and Microsoft 365 Copilot, these integrations have greatly simplified the AI procurement process. There is no need to sign separate contracts with multiple vendors or separate settlements, significantly shortening the deployment cycle.
Claude's deployment in Microsoft Foundry uses serverless technology and is operated and maintained by Anthropic. Developers can quickly go online through the Foundry platform API and tools:
Start building now: Supports Foundry’s multiple APIs, development tools, and workflows.
Seamless connection with Azure agreement: applicable to Azure Consumption Commitment (MACC), directly integrated into the existing accounting system without separate approval.
Multi-language access: supports Python, TypeScript and C# SDK, and is compatible with Microsoft Entra certification.
At present, the Claude model has been officially launched through global standard deployment and is billed at standard API prices. The local data area in the United States will be launched soon. Specific prices and plans can be found on Anthropic’s official website.
For different scenarios, Microsoft customers can flexibly choose Claude cutting-edge models:
Sonnet 4.5: The world's leading code and agent model, suitable for high-end tasks such as complex reasoning, multi-step intelligent workflows, and autonomous programming.
Haiku 4.5: The fastest and one-third the cost of Sonnet. It is suitable for high-concurrency applications such as sub-agents, customer service automation, content review and real-time programming support.
Opus 4.1: A powerful professional reasoning model, good at dealing with high-complexity, multi-step problems and rigorous and detailed tasks.
All the above models can support Claude developer tools on the Foundry platform, such as code execution, web page retrieval, online crawling, intelligent citation, visual analysis, tool calling, prompt caching and more. Developers can check the official documentation for complete features.
Currently, Claude is open for public testing in Microsoft Foundry. Users can visit the Foundry directory to deploy Sonnet 4.5, Haiku 4.5 or Opus 4.1 models, or refer to the official documentation for in-depth details.