Asserting Anthropic’s latest mannequin, Claude Opus 4.5, in Microsoft Foundry. Opus 4.5 is now out there in public preview in Microsoft Foundry, GitHub Copilot paid plans, and Microsoft Copilot Studio.
We’re at an actual inflection level within the AI panorama, a threshold the place fashions transfer from helpful assistants to real collaborators. Fashions that perceive the target, consider constraints, and execute advanced multi-tool workflows. Fashions that not solely help processes, however assist restructure them for reliability, scale, and operational effectivity.
Anthropic’s latest mannequin, Claude Opus 4.5, embodies that shift. At present, we’re excited to share that Opus 4.5 is now out there in public preview in Microsoft Foundry, GitHub Copilot paid plans, and Microsoft Copilot Studio.
Constructing on the Microsoft Ignite announcement of our expanded partnership with Anthropic, Microsoft Foundry delivers its dedication to giving Azure prospects instant entry to the widest choice of superior and frontier AI fashions of any cloud. Foundry empowers builders to speed up innovation with an built-in, interoperable, and safe AI platform that allows seamless deployment, integration, and scaling for AI apps and brokers.
We’re excited to make use of Anthropic Claude fashions from Microsoft Foundry. Having Claude’s superior reasoning alongside GPT fashions in a single platform provides us flexibility to construct scalable, enterprise-grade workflows that transfer far past prototypes.
—Michele Catasta, President, Replit
Opus 4.5 for actual work
Opus 4.5 units a brand new bar for coding, agentic workflows, and enterprise productiveness: outperforming Sonnet 4.5 and Opus 4.1, at a extra accessible value level. Its versatility throughout software program engineering, advanced reasoning, software use, and imaginative and prescient unlocks new alternatives for organizations to modernize programs, automate crucial workstreams, and ship ROI quicker.
By prioritizing fast integration of the most recent fashions, Foundry permits Azure prospects to remain forward of the curve and maximize the impression of their agentic AI programs; all whereas sustaining centralized governance, safety, and observability at scale.
1. Constructed for manufacturing engineering and agentic capabilities
In line with Anthropic, Opus 4.5 delivers state-of-the-art efficiency on trade normal software program engineering benchmarks, together with new highs on SWE-bench (80.9%). Early testers constantly describe the mannequin as with the ability to interpret ambiguous necessities, purpose over architectural tradeoffs, and determine fixes for points that span a number of programs.
Opus 4.5 accelerates engineering velocity by finishing multi-day growth work in hours with:
- Improved multilingual coding efficiency
- Extra environment friendly code era
- Stronger check protection
- Cleaner architectural and refactoring selections
| Functionality / Benchmark | Claude Opus 4.5 | Claude Sonnet 4.5 | Claude Opus 4.1 | Gemini 3 Professional |
| Agentic coding (SWE-bench Verified) | 80.90% | 77.20% | 74.50% | 76.20% |
| Agentic terminal coding (Terminal-bench 2.0) | 59.30% | 50.00% | 46.50% | 54.20% |
| Agentic software use — Retail (t2-bench) | 88.90% | 86.20% | 86.80% | 85.30% |
| Agentic software use — Telecom (t2-bench) | 98.20% | 98.00% | 71.50% | 98.00% |
| Scaled software use (MCP Atlas) | 62.30% | 43.80% | 40.90% | _ |
| Pc use (OSWorld) | 66.30% | 61.40% | 44.40% | _ |
| Novel downside fixing (ARC-AGI-2 Verified) | 37.60% | 13.60% | _ | 31.10% |
| Graduate-level reasoning (GPQA Diamond) | 87.00% | 83.40% | 81.00% | 91.90% |
| Visible reasoning (MMMU validation) | 80.70% | 77.80% | 77.10% | _ |
| Multilingual Q&A (MMLU) | 90.80% | 89.10% | 89.50% | 91.80% |
Claude Opus 4.5 benchmark outcomes from Anthropic
Opus 4.5 can also be one of many strongest tool-using fashions out there at present, able to powering brokers that work seamlessly throughout tons of of instruments. Builders achieve entry to a number of vital upgrades:
- Programmatic Device Calling: Execute instruments instantly in Python for extra environment friendly, deterministic workflows.
- Device Search: Dynamically uncover instruments from massive libraries with out utilizing up house within the context window.
- Device Use Examples: Extra correct software calling for advanced software schemas.
Collectively, these capabilities allow refined brokers throughout cybersecurity, full-stack software program engineering, monetary modeling, and different workflows requiring a number of software interactions. Opus 4.5 exhibits sturdy, real-world intelligence making use of these instruments creatively inside constraints. In testing, the mannequin efficiently navigated advanced coverage environments, corresponding to airline change guidelines, chaining upgrades, downgrades, cancellations, and rebookings to optimize outcomes. This type of adaptive, constraint-aware problem-solving displays a significant step ahead in what agentic AI programs can accomplish.
Manus deeply makes use of Anthropic’s Claude fashions due to their sturdy capabilities in coding and long-horizon process planning, along with their prowess to deal with agentic duties. We’re very excited to be utilizing them now on Microsoft Foundry!
—Tao Zhang, Co-founder & Chief Product Officer, Manus AI
2. Improved developer expertise on Foundry
Opus 4.5 paired with new developer capabilities provided on Foundry is designed to assist groups construct more practical and environment friendly agentic programs:
- Effort Parameter (Beta): Management how a lot computational effort Claude allocates throughout pondering, software calls, and responses to steadiness efficiency with latency and price on your particular use circumstances.
- Compaction Management: Deal with long-running agentic duties extra successfully with new SDK helpers that handle context effectively over prolonged interactions.
These enhancements present higher predictability and operational management for enterprise workloads.
3. Enhanced workplace productiveness and laptop use
Opus 4.5 additionally doubles down as Anthropic’s finest imaginative and prescient mannequin, unlocking workflows that rely on advanced visible interpretation and multi-step navigation. Pc use efficiency has improved considerably, enabling extra dependable automation of desktop duties.
For data staff, the mannequin delivers a step-change enchancment in powering brokers that create spreadsheets, shows, and paperwork. It produces work with consistency, skilled polish, and real area consciousness making it a match for finance, authorized, and different precision-critical verticals. The mannequin higher leverages reminiscence to keep up context and consistency throughout recordsdata all through sprawling skilled initiatives.
4. Security and safety
In line with Anthropic, Opus 4.5 additionally delivers significant enhancements in security and safety. The mannequin exhibits a decreased fee of misaligned responses, stronger robustness in opposition to prompt-injection assaults, and extra dependable conduct throughout advanced duties.
These enhancements align with Microsoft’s dedication to offering enterprise prospects with fashions that meet excessive bars for security, governance, and operational integrity
Use circumstances
Opus 4.5 serves the next use circumstances
- Software program growth: Deploy brokers that deal with advanced, multi-system growth duties with minimal supervision.
- Monetary evaluation: Join insights throughout regulatory filings, market reviews, and inner information for stylish predictive modeling and proactive compliance monitoring.
- Cybersecurity: Correlate logs, vulnerability databases, and risk intelligence for professional-grade risk detection and automatic incident response.
- Enterprise operations: Handle refined workflows requiring coordination throughout a number of instruments, programs, and knowledge sources.
Pricing and availability
Opus 4.5 delivers frontier efficiency and units a brand new normal for a wide range of use circumstances at one third the value of earlier Opus-class fashions.
|
Mannequin |
Provide sort |
Deployment sort |
Areas |
Worth (1M tokens) |
Availability |
|
Claude Opus 4.5 |
Serverless Pay-go |
World Commonplace |
East US2, Sweden Central |
Enter- $5 Output- $25 |
November 24, 2025 (public preview) |
Get began at present
Claude Opus 4.5 is out there now in Microsoft Foundry and coming quickly in Visible Studio Code through the Foundry extension. Go to the Foundry portal to start constructing with Opus 4.5.
