The catalog in Microsoft Foundry for discovering, comparing, testing and deploying models from Microsoft and partners such as OpenAI, Meta, Mistral and Cohere.
The Foundry model catalog is the part of Microsoft Foundry where models for generative AI applications are discovered, evaluated and deployed. It brings together foundation, reasoning, small language, multimodal, domain-specific and industry models in one place. The catalog lists models from Microsoft and from partners such as OpenAI, Meta, Mistral and Cohere, so that teams can work with a range of providers through the same portal instead of handling each model source separately.
According to Microsoft Learn, the catalog is organized into two categories. Models sold by Azure are hosted and sold by Microsoft under Microsoft Product Terms. They come with Microsoft support, enterprise SLAs and a responsible AI review. Models from partners and community come from third parties, research labs and other contributors. These are billed through the Azure Marketplace and are supported by their providers rather than by Microsoft. This split matters in practice, because it determines who handles support, how the model is billed and which terms apply.
Finding a suitable model is the first step. Users can search and filter the catalog by collection, region, deployment option, lifecycle stage, supported features and inference task. For comparison, the catalog offers a leaderboard and benchmarks, and each model has a model card. The model card is organized into tabs for quick facts, details, deployments, benchmarks and license, which gives a compact view of what a model offers and under which conditions it can be used.
Deployment follows one of two main options. Managed compute runs the model on dedicated virtual machines and is billed by VM core hours; it requires VM quota and a hub-based project. Serverless deployment exposes the model through a Microsoft-hosted API and is billed mostly by tokens. Choosing between them is a typical decision: dedicated capacity versus pay-per-use consumption, with different prerequisites. The documentation also notes that content filtering is not available for embedding and time-series models, which is relevant when planning responsible AI controls around a deployment.
Typical uses of models from the catalog include copilots, agents, retrieval-augmented generation and fine-tuning. The M365.FM episode "Build Auditable AI Agents with Azure AI Foundry" touches on this role. It lists the Foundry model catalog among the customization options for agents: developers can select from the ecosystem of models available there to fit different industry needs. The same overview names further options, including grounding agents with knowledge from Bing, SharePoint, Microsoft Fabric and Azure AI Search, over 1,400 action connectors through Azure Logic Apps, data upload options, and API and function calls. In that picture, the model catalog supplies the model choice, while the other options shape what the agent knows and can do.
In summary, the Foundry model catalog serves as the entry point for model selection in Microsoft Foundry. It combines discovery, comparison, testing and deployment, and it distinguishes clearly between Microsoft-supported models and partner or community models with their own support and billing arrangements.
Overview compiled with AI from M365.FM episodes and the facts on this page. You can improve it with the edit button below.
In this episode, we unpack two major AI platforms in the Microsoft ecosystem—Copilot Studio and Azure AI Foundry—and help listeners understand when to use each one. Whether you're building conversational AI, custom machine learning models, or enterprise-grade…
Frequently asked questions
Questions marked with AI were written with AI from the facts on this page.
What is the Foundry model catalog? (AI)
It is the hub in Microsoft Foundry for discovering, evaluating and deploying models for generative AI applications. It lists foundation, reasoning, small language, multimodal, domain-specific and industry models from Microsoft and partners.
What is the difference between models sold by Azure and models from partners and community? (AI)
Models sold by Azure are hosted and sold by Microsoft under Microsoft Product Terms, with Microsoft support, enterprise SLAs and a responsible AI review. Partner and community models are billed through the Azure Marketplace and supported by their providers rather than Microsoft.
How can models from the catalog be deployed? (AI)
Either as managed compute on dedicated virtual machines, billed by VM core hours, or as serverless deployment through a Microsoft-hosted API, billed mostly by tokens. Managed compute needs VM quota and a hub-based project.
Short info above, community knowledge below: add practical tips, examples or corrections. New contributions are reviewed first; trusted members publish directly.
Experts and companies for this topic
Nobody has listed this topic yet. Add it to your profile.