Microsoft-Decision-1 is Microsoft's decision model: given a fixed set of possible answers, it gives each one a probability instead of writing text. It is offered in Microsoft Foundry and on OpenRouter.
In plain words
Most language models write free text. A decision model instead chooses among answers you supply, which suits the many small choices inside apps and agents. Microsoft-Decision-1 gives every option a calibrated probability, so software can act when it is sure and ask a person when it is not. It handles yes/no, multiple-choice and rating questions, and can grade AI answers against a rubric. Microsoft built it from Qwen3.5-9B and says it beat LLMs and other decision models in its tests. It was released on 9 October 2026.
A quick example
Imagine a help-desk app that must send each new ticket to billing, sales or tech support. It passes the ticket and the three team names to a decision model, gets back a probability for each team, and routes the ticket on its own only when one team scores very high. Otherwise a person checks it.
Where you'll meet it
- Routing a request to the right model, tool or team inside an agent.
- Labelling data and grading AI answers against a rubric.
Related
Sources
- DocsIntroducing Microsoft-Decision-1, our model for fast decision-making, Microsoft (Command Line) · read 9 Oct 2026
- DocsMicrosoft-Decision-1, OpenRouter · read 9 Oct 2026
Checked by WhatIsAI against the sources above · 9 Oct 2026. A full five-minute explainer with a diagram may follow. Spotted an error? Report it.