AI models are increasingly taking on work that extends far beyond a single prompt: building a feature across a codebase, investigating a complex issue, synthesizing hundreds of pages of information, or working through a multi-step business process. As that work gets longer, raw intelligence is only part of what matters. The model also needs to stay focused, make good decisions along the way, communicate what it is doing, and produce work that people can quickly review and use.
Today, Claude Opus 5. 5 is available in Microsoft Foundry, bringing Anthropic’s most capable Opus model to developers and enterprises building AI applications and agents. Claude Opus 5.
5 is designed for everyday complex work. It advances Opus 5 across agentic coding, knowledge work, and long-running tasks while making it easier for people to understand what the model did, what it found, and what it needs next. Claude Opus 5.
5 also does more with fewer tokens. Lower per-token prices and much cheaper cache reads stack on top of the efficiency gains. Built for work that takes time Writing a function is one thing.
Building a feature that touches multiple services, tracing a production issue across a large repository, or carrying a task from investigation through implementation and validation is something else entirely. Claude Opus 5. 5 is designed for these longer-running workflows.
For software development, it can work through long-running coding tasks such as building features across a codebase, debugging, refactoring, and reviewing code. It finds the root cause before changing anything, checks its work as it goes, and explains its changes in plain language, so engineers can review and trust them quickly. That combination becomes particularly valuable when developers use models through agentic coding environments, where a session may involve dozens of steps and run for an extended period of time.
The same applies beyond software development. For knowledge workers, Claude Opus 5. 5 can bring together information from multiple sources, work through long documents and spreadsheets, perform analysis, and help create artifacts such as memos, reports, and presentations.
Compared with Opus 5, it produces outputs that require less editing before they are ready to share. An AI model that communicates more like a teammate As agents take on more autonomous work, another challenge emerges: keeping the human in the loop without overwhelming them. An agent that performs 50 steps should not require someone to inspect 50 steps to understand whether the work was successful.
Claude Opus 5. 5 introduces improvements to agentic communication designed to make long-running work easier to follow. As it works, the model can surface the information that matters most: What it did What it found What decisions it made Where it needs input from the user What happened at the end of a long-running task The goal is simple: spend less time decoding what the model did and more time using the result.
This matters particularly for enterprise agents, where users need to understand not only the final answer but also when an agent needs clarification, encounters a constraint, or reaches a decision point that requires human judgment. Adpative thinking Claude Opus 5. 5 uses adaptive thinking, automatically determining how much reasoning a task requires.
Rather than turning thinking on or off or manually specifying a thinking-token budget, developers use effort to influence how much work the model should put into a request. This allows the model to adapt its reasoning to the task at hand—from relatively straightforward requests to complex problems that require deeper analysis. For developers building agents, this can reduce the amount of application logic needed to decide when and how a model should reason.
Designed for long-running agent architectures Long-running agents create challenges beyond model intelligence. Conversations can exceed context limits. Tools available to an agent can change.
Applications may need to compact earlier context while preserving the model’s understanding of the work already completed. Alongside Claude Opus 5. 5, Anthropic is introducing beta API capabilities designed for these scenarios, including asynchronous compaction, keep-tail compaction, and changing tools during a conversation while preserving thinking and prompt caching.
These capabilities can help agent developers maintain continuity across longer tasks without rebuilding the state of the application every time context or available tools change. Combined with Microsoft Foundry, developers can use Claude Opus 5. 5 as part of broader agent systems that connect models with enterprise data, tools, evaluation, and operational workflows.
Expanded safeguards for more capable models As model capabilities increase, Anthropic is also expanding the safeguards applied to Claude Opus 5. 5. Claude Opus 5.
5 is the first Opus model to use safety classifiers like those introduced with Claude Fable 5. 1 in areas including cybersecurity, biology, AI development, and distillation. For common developer, educational, and knowledge-work scenarios, customers can continue using the model for tasks such as identifying software vulnerabilities or learning about biological concepts.
Certain requests that Anthropic identifies as higher-risk or dual-use may be handled by another Claude model with the appropriate safeguards. This reflects an increasingly important part of deploying more capable models: advancing what models can do while applying safeguards appropriate to the capabilities they introduce. Pricing Model Deployment Offers Input/M Tokens Output/M Tokens Availability Claude Opus 5.
5 Global Standard, US DataZone $4Cache Hit – $0. 20Cache Write – $5Cache Write (1 hr) – $8 $20 GA, Hosted on Azure Claude Opus 5. 5 (Long Context) Global Standard, US DataZone $4Cache Hit – $0.
20Cache Write – $5Cache Write (1 hr) – $8 $20 GA, Hosted on Azure Build with Claude Opus 5.
Originally published at techcommunity.microsoft.com
