Microsoft Foundry の新機能 | 2026年6月What’s New in Microsoft Foundry | June 2026

Microsoft has published its June 2026 update for Microsoft Foundry, the company's platform for building, evaluating, and deploying AI applications and agents. The monthly release rolls together new model availability, developer tooling enhancements, and platform refinements that, according to the announcement, are aimed at helping teams move from prototype to production more quickly. For organizations already standardizing on Azure for their AI workloads, these incremental updates matter because they shape the day-to-day workflow of prompt engineering, model selection, and application deployment.

The most visible change in monthly Foundry releases is typically the expansion of the model catalog, and the June update appears to continue that pattern. Foundry maintains a broad selection of models from Microsoft, OpenAI, and third-party and open-source providers, allowing developers to compare options and switch between them without rebuilding their applications. Adding newer model versions and specialized variants gives teams more choice around cost, latency, context window size, and multimodal capability. The practical benefit is that a developer can often adopt a more capable or more efficient model by changing a deployment configuration rather than re-architecting an application, which lowers the friction of keeping pace with a fast-moving field.

On the developer tooling side, the update is described as strengthening the workflows used to build and ship AI features. Foundry combines a web portal with SDKs and integration into familiar environments such as Visual Studio Code, and improvements here generally target the loop of testing prompts, wiring up retrieval-augmented generation, and connecting external data and tools. Enhancements to the software development kit and the underlying APIs tend to make it easier to manage deployments programmatically, integrate Foundry into continuous integration pipelines, and reduce the amount of custom code teams must maintain. These changes are incremental by design, but collectively they reflect Microsoft's effort to make the platform approachable for mainstream developers rather than only specialists.

Agent development is an area Microsoft has emphasized heavily across recent Foundry releases, and it is likely that this update touches the Foundry Agent Service or related capabilities. Agents—systems that can call tools, retrieve information, and take multi-step actions—require orchestration, memory, and connections to enterprise data. Foundry positions itself as a place to build such agents with built-in support for grounding responses in an organization's own content and for interoperating with the broader Microsoft ecosystem, including Copilot experiences and Azure services. Improvements to observability, tracing, and evaluation are commonly part of this story, because teams deploying agents in production need to understand why a model produced a given output and to catch regressions before they reach users.

Governance and responsible AI features are another recurring theme. Foundry includes evaluation tooling, content safety controls, and monitoring intended to help enterprises meet compliance and reliability requirements. As AI systems take on more consequential tasks, the ability to measure quality, detect harmful content, and audit behavior becomes a prerequisite for real deployments rather than an optional add-on. Updates that refine these guardrails are consistent with Microsoft's stated approach of pairing capability with controls.

It helps to place Foundry in context. The platform evolved from what Microsoft previously branded as Azure AI Studio and Azure AI Foundry, reflecting a consolidation of the company's AI development tools under a single name. This sits alongside adjacent offerings such as Azure OpenAI Service, Copilot Studio for lower-code agent building, and the Microsoft 365 Copilot family aimed at end users. The competitive backdrop includes comparable platforms from Google, through Vertex AI, and from Amazon, through Bedrock, as well as independent tooling from companies focused on orchestration, evaluation, and vector search. Each provider is racing to offer a unified environment that spans model access, application frameworks, and operational monitoring.

For teams evaluating these announcements, the sensible approach is to review the official Foundry documentation and release notes for the precise list of models, regions, and features, since availability often varies by geography and pricing tier and some capabilities launch in preview before reaching general availability. The June 2026 update, taken as a whole, appears less about a single headline feature and more about the steady accumulation of improvements that make an AI platform dependable. That cadence, rather than any one release, is what ultimately determines whether developers can build and operate AI applications with confidence.