What is Foundry Models?
Foundry Models is the central model catalog in Microsoft Foundry (formerly Azure AI Foundry). It provides access to a very large number of foundation models: models sold and operated directly by Microsoft (“Models sold by Azure”), as well as models from partners and the community such as Anthropic, Meta, Mistral, DeepSeek, and Hugging Face.
Core Features
- Model Discovery: Searchable catalog with filters by provider, task, industry, and capabilities (e.g., reasoning, tool calling)
- Benchmarks and Comparison: Leaderboard and benchmark metrics for select models for direct comparison
- Serverless Deployment: Use models as an API with token-based billing, without managing your own infrastructure
- Managed Compute: Deploy model weights on dedicated VMs for full control, built on Azure Machine Learning
- Fine-tuning: Adapt select models to domain-specific data, via managed compute or serverless deployment depending on the model
Typical Use Cases
LLM Selection for Enterprise: Companies compare models like GPT, Claude, Llama, and Mistral directly in Microsoft Foundry. The catalog shows benchmark results and deployment options for informed model selection.
Open-Source Strategy: Teams committed to open models find numerous Hugging Face models as well as open model families such as Llama or Mistral. Depending on licensing, these can run on Azure infrastructure.
Specialized Use Cases: For image analysis, reasoning, or industry-specific tasks, the catalog offers specialized models with their own benchmark and licensing details.
Benefits
- Unified access to models from numerous providers through one platform
- Clear distinction between Microsoft-sold models with enterprise support and partner/community models
- Transparent deployment options (serverless vs. managed compute) with model-dependent pricing shown before deployment
- Integration with Azure Machine Learning and the broader Microsoft Foundry ecosystem (fine-tuning, observability, responsible AI)
Integration with innFactory
As a Microsoft Solutions Partner, innFactory supports you with Foundry Models: we help with model selection for your use case, implement deployment pipelines for serverless or managed-compute scenarios, and optimize costs and governance across the model lifecycle.
Typical Use Cases
Frequently Asked Questions
Which models are available in the Foundry model catalog?
The catalog contains thousands of models from Microsoft, OpenAI, Anthropic (Claude), Meta (Llama), Mistral, DeepSeek, Hugging Face, and other providers. From LLMs to embedding models to vision and industry-specific models.
What is the difference between Managed Compute and Serverless?
Managed Compute deploys model weights on dedicated VMs, billed by compute time. Serverless deployments provide a Microsoft-hosted API, billed per token or usage, without you managing infrastructure.
Can I use open-source models?
Yes, the catalog contains numerous open models, including from Hugging Face. Depending on the provider, these can run on Azure infrastructure via managed compute or serverless deployment.
How do I choose the right model?
The catalog offers filters by provider, task, and capabilities, along with benchmark data and a leaderboard for select models. Microsoft Foundry enables direct comparison and testing before production use.
Are the models suitable for enterprise use?
Models sold directly by Microsoft ('Models sold by Azure') offer deep Azure integration and enterprise SLAs. Models from partners and the community are supported by their respective providers, with varying support and SLA terms.
Note: All product information on this page has been compiled with care, but is provided without guarantee and may be outdated or incomplete. Cloud services evolve rapidly — features, pricing, SLAs, and availability change frequently. Authoritative and up-to-date information can only be found on the official product page of Azure (official documentation). This page does not represent an offer by Azure.
