Microsoft releases a model that scores choices instead of writing answers
Microsoft-Decision-1 makes fixed-choice decisions at a listed $0.042 per million input tokens, creating a low-cost option to test for routing and classification.
Microsoft’s October 9 release targets the small judgments inside larger workflows: which queue receives a request, whether an answer meets a rubric, or when a case needs review.
Give Decision-1 the content, a question and defined choices; it returns scores for those choices. It is not a conversational assistant. [1][2]
The model is post-trained from Qwen3.5-9B. Microsoft’s Foundry post describes a public preview, and OpenRouter now lists the model with Azure as its provider. Both Microsoft and OpenRouter list $0.042 per million input tokens and no output-token charge. [1][2][3]
The useful experiment is a bounded one: replay previously resolved cases, measure wrong routes and cases deferred to people, and include the whole workflow’s cost. Probability scores need to be checked against observed outcomes in your own setting before they govern consequential actions. Microsoft’s benchmark rankings do not establish your error rate.
Take it further
Explore the model Public catalog is readable without sign-in; API use requires an account and billing. No trial credit, regional entitlement or deadline is promised.
Sources & details 3 sources
Discuss in your voice app
Take the reporting, visual explanations and original sources into a conversation. Copy the brief, paste it into your preferred assistant, then start its voice mode.
Paste the brief into your voice app to discuss this reporting.Reporting sources
- Microsoft · October 9 release
Original release, base model, structured-choice contract and list price. Vendor benchmark superiority is not adopted as a demonstrated result.
- OpenRouter · Live provider and price listing
Separately operated distribution catalog checked October 10: model, Azure provider, $0.042/M input and $0/M output. Listing is observable; no paid call or throughput test was made.
- Microsoft Foundry · Public-preview scope
Original October 9 launch text was retrieved in search results; a later direct open failed. It explicitly says public preview. No universal account eligibility or region availability is inferred.
Images & figures
AI-generated editorial illustration. Invented physical metaphor for scoring supplied choices, not the model implementation, an actual ranking, numerical probabilities, accuracy or speed.
Euangelion · AI-generated editorial illustration · Source · Usage terms
Defined choices make a decision inspectable. Permission still belongs to the application.
Euangelion · original explanatory figure · Source · Usage terms · Download figure