What is Foundry Models?
Foundry Models is a central hub for AI models in Microsoft Foundry. The Catalog provides access to hundreds of foundation models from Microsoft, OpenAI, Meta, Mistral, Hugging Face, and other providers for various AI use cases.
Core Features
- Model Discovery: Searchable catalog with filters by task, provider, and license
- Benchmarks and Comparison: Performance metrics and direct model comparison for informed decisions
- Serverless Deployment: Pay-per-token without infrastructure for quick start
- Managed Compute: Dedicated VMs for full control and customization
- Fine-tuning: Adapt models to domain-specific requirements
Typical Use Cases
LLM Selection for Enterprise: Companies compare GPT-4, Claude, Llama, and Mistral directly in Azure. The Catalog shows costs, latency, and benchmark results for informed model selection.
Open-Source Strategy: Teams committed to open-source models find Llama, Falcon, and hundreds of Hugging Face models. These run on Azure infrastructure with enterprise-grade security.
Specialized Computer Vision: For image analysis, the Catalog offers Florence, CLIP, and other vision models. From object detection to image generation, specialized models are available.
Benefits
- Unified access to models from various providers
- Enterprise security and compliance for all models
- Transparent pricing and benchmark data
- Seamless integration with Azure AI Foundry and Azure ML
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, and optimize costs through the right choice between Serverless and Managed Compute.
Typical Use Cases
Frequently Asked Questions
Which models are available in the Catalog?
The Catalog contains models from OpenAI (GPT-4), Meta (Llama), Mistral, Hugging Face, Microsoft, and more. From LLMs to embedding models to computer vision.
What is the difference between Managed Compute and Serverless?
Managed Compute deploys models on dedicated VMs with full control. Serverless offers pay-per-token without infrastructure management for quick start.
Can I use open-source models?
Yes, the Catalog contains hundreds of open-source models from Hugging Face. These can be deployed on Azure infrastructure.
How do I choose the right model?
The Catalog shows benchmarks, use cases, and pricing for each model. Azure AI Foundry enables direct comparison and testing before deployment.
Are the models suitable for enterprise use?
Yes, all models in the Catalog run in the secure Azure environment. Data protection, compliance, and SLAs apply as with other Azure services.
