Adopt
Microsoft Foundry (formerly Azure AI Foundry) is Microsoft's unified platform for building AI applications and agents on Azure. It replaces the standalone "Azure OpenAI Service" branding - OpenAI models are now consumed as "Azure OpenAI in Microsoft Foundry Models".
Key Features
- Model catalog: OpenAI frontier models plus open-weight and third-party models (Meta, Mistral, DeepSeek, xAI and others) behind one API.
- Agent tooling: Foundry Agent Service for hosting agents, with built-in evaluations, guardrails/content safety and observability.
- Enterprise integration: Entra ID, private networking, regional data residency and compliance certifications - the same reasons Azure OpenAI was our enterprise default.
Recommendation
For teams already in the Microsoft/Azure ecosystem, Foundry is the default way to consume OpenAI models with enterprise requirements. Note that Azure retires models on fixed schedules (e.g. the GPT-4.1 family on 2026-10-14) - plan model upgrades as recurring maintenance, pin API versions and test before switching defaults (retirement schedule).
Alternatives: AWS Bedrock for multi-model access in the AWS ecosystem, Hugging Face or Replicate.ai for lower onboarding hurdles.