Microsoft has announced two major upgrades for the Microsoft Foundry Models ecosystem. The company is expanding its intelligent model router with seven new AI models while also launching a new open-source auto-evaluation repository for developers and enterprises building AI applications at scale.
The update strengthens Microsoft’s push toward multi-model AI workflows, giving developers more flexibility to optimize applications for reasoning, speed, cost, and performance.
7 New Models Added to Foundry Model Router
The Microsoft Foundry model router now supports seven additional AI models:
- GPT-5.4
- GPT-5.4-mini
- GPT-5.4-nano
- GPT-5.3-chat
- Claude-Opus-4.7
- GPT-5.5
- Grok-4.1-fast-reasoning
These models join the growing catalog available through Microsoft’s AI infrastructure and can be automatically selected by the router depending on workload requirements.
Microsoft’s model router is designed to intelligently analyze prompts and dynamically route requests to the best-performing large language model based on task complexity, latency, reasoning requirements, and cost efficiency.
This means developers no longer need to manually decide which AI model should handle every request. Instead, the system can automatically optimize for performance and pricing in real time.
Why This Matters for Developers
The addition of newer GPT-5.x models alongside Anthropic’s Claude and xAI’s Grok models highlights Microsoft’s increasingly open multi-model AI strategy.
Developers building on Microsoft Azure AI Foundry can now mix and match models from multiple providers while using the router as an orchestration layer.
Key advantages include:
- Better reasoning for complex AI agents
- Faster responses for lightweight tasks
- Lower operational costs through smart routing
- Improved scalability for enterprise AI apps
- Reduced latency without sacrificing output quality
Microsoft previously said its model router could reduce costs significantly compared to directly using larger frontier models while maintaining similar accuracy levels.
Open-Source Auto-Evaluation Repository Also Launches
Alongside the model expansion, Microsoft is making a new open-source auto-evaluation repository available for developers.
The repository is aimed at helping teams benchmark, evaluate, and compare AI model performance across different workloads and use cases. While Microsoft has not yet detailed every feature publicly, the move aligns with the industry-wide push toward transparent AI testing and reproducible evaluations.
The launch also complements Microsoft’s broader investments in AI benchmarking, agentic workflows, and automated model selection tools within Foundry.
Microsoft Is Doubling Down on the “Open Agentic Web”
The announcement arrives as Microsoft continues expanding its AI ecosystem during the Microsoft Build 2025 era of AI agents.
At Build, Microsoft emphasized that enterprises are increasingly adopting multi-model AI strategies and autonomous AI agents powered by advanced reasoning systems.
By integrating OpenAI, Anthropic, xAI, and other frontier models into a single routing platform, Microsoft is positioning Foundry as a central hub for enterprise AI development.
The combination of:
- intelligent routing,
- broader model choice,
- open-source evaluation tooling,
- and scalable Azure infrastructure
could make Foundry increasingly attractive for organizations building next-generation AI assistants, copilots, and autonomous agents.
Availability
The newly added models are rolling out through Microsoft Foundry’s model router infrastructure starting today. Developers can explore supported models and deployment options through the official Microsoft Foundry Model Catalog.
