Microsoft-Decision-1

Direct from Azure models are a select portfolio curated for their market-differentiated capabilities:

• Secure and managed by Microsoft: Purchase and manage models directly through Azure with a single license, consistent support, and no third-party dependencies, backed by Azure's enterprise-grade infrastructure.
• Streamlined operations: Benefit from unified billing, governance, and seamless PTU portability across models hosted on Azure - all as part of one Microsoft Foundry platform.
• Future-ready flexibility: Access the latest models as they become available, and easily test, deploy, or switch between them within Microsoft Foundry; reducing integration effort.
• Cost control and optimization: Scale on demand with pay-as-you-go flexibility or reserve PTUs for predictable performance and savings.

Learn more about Direct from Azure models .

Microsoft-Decision-1 is a decision-scoring model: given a situation and a question with a fixed set of answer options, it returns a calibrated probability for each option instead of generated text. It is built on Alibaba's open-weight Qwen3.5-9B and post-trained by Microsoft. Training data consists of publicly available datasets subject to Microsoft's Open Data process, plus synthetic data created by the team. We will also rebase on other models, including MAI and OpenAI.

• Calibrated decision scoring: Returns a calibrated probability score for every supplied answer option.
• Flexible question formats: Supports yes/no, multiple-choice, rating, classification, and rubric-based questions.
• AI and agent evaluation: Grades AI responses and proposed agent actions against supplied criteria or rubrics.
• Classification and routing: Supports request classification, relevance judgment, triage, and routing workflows.
• Groundedness evaluation: Assesses responses against evidence provided in the input.
• Safety and guardrail checks: Supports content-safety flagging and application-defined thresholds.
• Explicit abstention: Supports options such as "cannot tell" when the supplied evidence is insufficient.
• Single-pass scoring: Produces structured decision scores in a single model invocation over inputs of up to 32K tokens.

• AI-output evaluation: Grade AI-generated responses against a defined rubric or evaluate whether a response is grounded in supplied evidence.
• Classification and routing: Classify requests, evaluate relevance, route requests, and triage workflow items.
• Agent guardrails: Evaluate proposed agent actions or tool calls before an integrating application permits execution.
• Content-safety screening: Flag potentially harmful content using application-defined probability thresholds.
• Search and document relevance: Judge whether search results or documents are relevant to a supplied question or task.
• Confidence-based automation: Automate high-confidence outcomes and escalate low-confidence or ambiguous outcomes for additional review.

Microsoft-Decision-1 is not designed for text generation, open-ended question answering, conversation, translation, or summarization. It is not intended for tasks that do not provide a closed question and a defined set of answer options, or for tasks that require knowledge not present in the input.

Microsoft-Decision-1 is not designed or evaluated for use as the sole automated decision-maker in consequential decisions about people. It should not be used as the sole basis for decisions involving credit, employment, housing, insurance, education, healthcare, legal rights, or similarly consequential domains.

The model should not be used for surveillance, profiling, or tracking individuals; suppression of lawful speech; or any use that violates applicable law, Microsoft's acceptable use policies, or applicable product terms.

The model is text-only and does not accept or produce image, audio, or video content. It does not generate explanations or rationales. The integrating application is responsible for defining appropriate answer options, confidence thresholds, escalation paths, human oversight, and safeguards for downstream actions.