Microsoft officially launched its decision model Microsoft-Decision-1.1 yesterday (the 17th), and it is now available on the Microsoft Foundry, OpenRouter, and Vercel platforms. It doesn’t do chit-chat; it focuses on high-frequency decisions in enterprise software: routing, classification, content verification, and workflow control. It only picks answers from permitted actions within rules set by developers, and it outputs calibrated confidence scores—when information is insufficient, it would rather “abstain” or prompt for human intervention instead of leaving hidden risks in critical business operations.

image.png

68 points, beating a host of rivals and twice as fast as the previous generation On the Artificial Analysis Decision Index v2.1 benchmark, it scored 68, up 6 points from the previous generation and ahead of competitors such as Jev 1.14, Google Gemini Decision 1, and OpenAI GPT-6 Luna. It is more than twice as fast as the previous generation, yet its price is as low as $0.042 per million input tokens, with output free. Microsoft also built a custom inference engine for it and optimized it based on a medium-sized mixture-of-experts model.

4,000 customers tested the waters first Since the public preview, more than 4,000 customers have used it, in scenarios including model routing, customer service ticket triage, and business process automation. Early users include Epic and Raiffeisen Bank. Microsoft also uses it internally to route requests for Copilot; its accuracy is comparable to large language models, but it is 2.2 times faster and 4.4 times cheaper. It also supports multimodal inputs such as photos and screenshots.