The Foundry Model Catalog is one of the platform's most powerful features — a curated marketplace of 1,800+ AI models from multiple providers, continuously updated with the latest releases.
| Provider | Key Models | Strengths |
|---|---|---|
| OpenAI | GPT-5.5, GPT-5.4, GPT-5.2, GPT-4o, o4-mini, o3 | Frontier reasoning, omnimodal, agentic tool-calling |
| Microsoft (MAI) | MAI-Transcribe-1, MAI-Voice-1, MAI-Image-2, Phi-4 | First-party speech/vision, efficient on-device |
| Microsoft Research | MagenticBrain, Fara1.5-9B | Cutting-edge research models for specialized reasoning and efficiency |
| Meta | Llama 3.3 70B, Llama 3.2 | Open-source, customizable |
| Mistral | Mistral Large, Ministral 3B | Efficient, multilingual |
| Alibaba | Qwen3 32B | Multilingual reasoning |
| xAI | Grok 4.3 | High-throughput reasoning, real-time knowledge |
| Fireworks AI | DeepSeek V4, DeepSeek V3.2, Kimi 2.6 | Ultra-fast open-weight inference |
| Cohere | Command R+, Embed v3 | Enterprise RAG, embeddings |
Every model has a Model Card containing benchmarks, license info, supported deployment types, pricing, and sample code.
GPT-5.5 became Generally Available on Azure Foundry on April 23, 2026. It is an omnimodal frontier model with a 1M context window, priced at $5 / $30 per MTok (input / output). GPT-5.5 is also available on Amazon Bedrock as of June 1, 2026, making it the first OpenAI model accessible across both major cloud providers simultaneously.
Foundry includes a Model Router that can automatically select the most appropriate model for a given prompt or workflow. This means your application can dynamically choose between GPT-5.5 for the most demanding tasks, GPT-5.4 for complex reasoning, or a smaller model like Phi-4 for simple tasks — optimizing cost and speed without code changes.