Executive Overview
The enterprise AI landscape is reaching a pivotal inflection point, transitioning from conversational experimentation toward autonomous, production-ready AI agents capable of executing complex business logic. Microsoft has unveiled a massive suite of updates now Generally Available (GA) on Microsoft Foundry, its end-to-end platform for building, running, governing, and distributing enterprise AI agents.
As organizations move away from fragmented, multi-vendor AI pipelines, Microsoft Foundry has emerged as a central nexus for enterprise deployment. Over 100,000 organizations are currently building on the platform, with market leaders including Adobe, Telefónica, and Tata Consultancy Services (TCS) actively operating agentic workflows in production environments.
This major operational milestone delivers on promises outlined at Microsoft Build: giving developers a unified environment to build agents in their native IDEs, run them on trusted, enterprise-grade cloud infrastructure, and deploy them across Microsoft 365 and enterprise apps without stitching together disconnected third-party tools.
Key announcements include:
- The immediate General Availability of OpenAI’s new GPT-5.6 model series (Sol, Terra, and Luna) across 28 global regions.
- The General Availability of the Asia-Pacific (APAC) Data Zone, enabling strict regional data processing and compliance.
- Deep integration with developer tools, including the GitHub Copilot SDK (GA), the Foundry Toolkit for VS Code, and native support for the Claude Agent SDK and Microsoft Agent Framework.
- Advanced token economics and governance features, including Agent Optimizer, Toolboxes, Model Router, and dedicated ROI tracking capabilities.
Detailed Chronology & Strategic Architecture
The journey to general availability represents a shift in how enterprise software architecture handles autonomous intelligence. Historically, deploying an AI agent required developers to manually wire together large language model APIs, vector stores, custom orchestration frameworks, access control layers, and telemetry systems.
Microsoft Foundry consolidates these disparate primitives into three strategic pillars designed to take agents from local development to global scale.
+-----------------------------------------------------------------------------------+
| MICROSOFT FOUNDRY |
+------------------------------------+----------------------------------------------+
| 1. BUILD | 2. OPERATE & DISTRIBUTE |
| • VS Code Toolkit & Skill | • Action-Oriented Agent Runtime |
| • Open Framework Support | • Microsoft 365 & Copilot Integration |
| • Multi-Model Hub (GPT-5.6 Series) | • Enterprise Identity & Context Controls |
+------------------------------------+----------------------------------------------+
| 3. GOVERN & OPTIMIZE |
| • Full-Lifecycle Observability | • Token Economics & Prompt Caching |
| • Model Router & PTU Spillover | • Enterprise ROI & Usage Analytics |
+-----------------------------------------------------------------------------------+Pillar 1: Build with Any Framework and Model
Development begins directly inside familiar environments like GitHub Copilot and Visual Studio Code (VS Code). Through the Foundry Toolkit for VS Code and the Foundry skill, deployment pipelines bridge the gap between local code and cloud execution.
Developers are no longer locked into proprietary ecosystems; Foundry provides native target runtimes whether teams build using:
- Microsoft Agent Framework
- GitHub Copilot SDK (Now Generally Available)
- Claude Agent SDK
Pillar 2: Action-Oriented, Context-Aware Runtimes
Running an agent in production requires state management, persistent memory across user sessions, identity integration, and tool access. The updated Foundry platform natively provides hosting environments where agents act on real-world events, carry short- and long-term context, and integrate securely into workplace communication channels like Microsoft 365.
Pillar 3: End-to-End Governance, Observability, and Optimization
Deploying an agent without full visibility creates operational risks. Foundry elevates trust from an individual developer task to a platform-level guarantee. The runtime incorporates tracing, logging, security controls, automated evaluation benchmarks, and fine-grained spending controls to ensure enterprise agents remain secure, compliant, and cost-effective.
Supporting Context & Metrics
OpenAI GPT-5.6 Series Deployment & Pricing
Central to this release is the General Availability of OpenAI’s GPT-5.6 series within Microsoft Foundry Models and Microsoft Foundry Agent Service. Rather than forcing every workload onto a single, high-cost model, the GPT-5.6 family introduces three distinct tiers tailored to varying levels of reasoning, latency, and operational expense:
- GPT-5.6 Sol: The premium frontier reasoning model engineered for highly complex multi-step analytical tasks, autonomous decision-making, and deep domain processing.
- GPT-5.6 Terra: A balanced workhorse model designed for high-throughput operational tasks requiring strong contextual awareness at an optimized price point.
- GPT-5.6 Luna: A fast, lightweight model optimized for low-latency, specialized sub-tasks, routing functions, and high-frequency real-time interactions.
To ensure immediate parity and scale, Microsoft is launching GPT-5.6 through Global Standard, Global Priority Processing (accessible across all 28 global Azure regions), Data Zones Standard, and Global Provisioned access on day one.
GPT-5.6 Enterprise Pricing Breakdown (Standard Global)
The following table reflects the global standard pricing matrix (USD per million tokens), incorporating the latest price optimizations announced by OpenAI:
| Model | Context Profile | Deployment Tier | Input ($/M) | Cached Input ($/M) | Cached Writes ($/M) | Output ($/M) |
|---|---|---|---|---|---|---|
| GPT-5.6 Sol | Short Context | Standard Global | $5.00 | $0.50 | $6.25 | $30.00 |
| GPT-5.6 Terra | Short Context | Standard Global | $2.00 | $0.20 | $2.50 | $12.00 |
| GPT-5.6 Luna | Short Context | Standard Global | $0.20 | $0.02 | $0.25 | $1.20 |
Note: Specialized pricing tiers, including Priority Processing and Data Zone rates, are available via direct Microsoft enterprise engagement.
Regional Sovereignty: Asia-Pacific (APAC) Data Zone GA
Alongside global model availability, Microsoft announced the General Availability of the Asia-Pacific (APAC) Data Zone for Microsoft Foundry.
Data residency and regulatory sovereignty remain significant hurdles for enterprises adopting public cloud AI. The APAC Data Zone guarantees that data processing and fine-tuning for frontier OpenAI models remain localized within designated Asia-Pacific boundary regions. This eliminates the need for enterprise architecture teams to engineer custom, multi-region failovers or run isolated infrastructure stacks just to comply with local financial and data protection laws.
Platform Infrastructure & Token Economics
As production workloads scale from internal pilots to millions of daily agent calls, managing compute consumption becomes critical. Microsoft Foundry integrates built-in token optimization mechanics designed to maximize efficiency:
- Toolboxes in Foundry: Dynamically strips context windows by sending only the exact function schemas and API tools required for a given execution step, preventing prompt bloat.
- Agent Optimizer: Continuously evaluates agent run history against custom metrics to automatically optimize prompts, tool parameters, and model selections.
- Model Router: Intelligently inspects incoming user queries and routes simple requests to low-cost models (such as Luna) while reserving intensive reasoning tasks for high-capacity models (Sol).
- Prompt Caching & PTU Spillover: Reduces compute overhead on repeated context injections while Provisioned Throughput Unit (PTU) spillover safeguards system uptime by routing burst traffic seamlessly without service degradation.
- Native ROI Analytics: Connects raw infrastructure expenditure directly with business metrics, allowing engineering and finance leads to measure agent value against consumption costs in real time.
Official Statements & Enterprise Adoption
The enterprise transition to Microsoft Foundry is showcased by deployments across finance, telecommunications, and professional services. Digital-native and legacy institutions alike are replacing fragmented internal frameworks with Foundry’s unified control plane.
Highlighting the impact on regulated industries, Hongsoo Kim, Chief Data and AI Officer (CDAO) at Viva Republica (Toss), emphasized the importance of localized data processing for financial institutions:
"As financial institutions adopt AI, responsible data handling becomes foundational to trust. Microsoft Foundry’s APAC Data Zone allows us to keep data processing regionally anchored while accessing advanced AI models at scale. This gives us the confidence to accelerate AI innovation responsibly and reinforces our ambition to be a leading AI-powered financial platform in Asia."
Similar patterns are emerging across major global organizations:
- Telefónica is leveraging hosted agent runtimes to automate complex customer network diagnostic workflows, significantly lowering mean time to resolution (MTTR).
- Adobe is utilizing Foundry’s platform capabilities to orchestrate multi-agent workflows that connect cloud services securely under enterprise identity standards.
- Tata Consultancy Services (TCS) is building and deploying sector-specific operational agents across its global client base, shifting client delivery cycles from months to days.
Future Outlook & Developer Roadmap
The General Availability of these updates signals a broader shift in enterprise software: the convergence of cloud infrastructure, developer tooling, and foundation models into a unified runtime layer for AI agents.
Microsoft’s commitment to supporting open frameworks—such as the Claude Agent SDK alongside its own native frameworks—positions Foundry as an open platform for autonomous software development.
Technical Resources & Ecosystem Onboarding
To accelerate enterprise enablement, Microsoft has published a comprehensive collection of developer workshops, lab environments, and technical documentation:
- Core Quickstarts: The updated Foundry Hosted Agent Quickstart provides a step-by-step path to setting up, evaluating, and deploying production-ready agents via the Azure Developer CLI (
azd). - Educational Curricula: A 12-lesson open-source guide, AI Agents for Beginners, is now live alongside dedicated learning modules like Develop AI Agents in Azure.
- Specialized Workshops: Developers can access hands-on repositories, including the Hosted Agents Workshop (.NET), the Foundry Toolkit for VS Code Lab, and the ZavaShop Supply Chain Workshop for real-world multi-agent supply chain orchestration.
- Technical Evaluation Guides: Practitioners can leverage Evaluating AI Agents: A Practical Guide with Microsoft Foundry to establish rigorous testing pipelines before shipping code to production.
As businesses pivot from generative AI prototypes to autonomous operational infrastructure, Microsoft Foundry provides the runtime, security boundary, and cost-governance layer necessary to operate enterprise AI agents safely at scale. All announced features are live and accessible globally in Microsoft Foundry.
