Ordalin
Submit a tool

Website-checked tool profile

Microsoft-Decision-1

Microsoft’s decision-scoring model that returns calibrated probabilities over fixed options for routing, classification and AI judging.

Sources checked

Visit website ↗
Pricing
Pay-per-token API pricing in US dollars: input tokens cost $0.042 per million, and output tokens are free. Official source 1 ↗
Available as
Available as an API model in Microsoft Foundry and through OpenRouter, with documentation on Microsoft Learn. Official source 1 ↗
Primary group
Automation & Agents

What it does

Microsoft-Decision-1 is a decision-scoring model from Microsoft. Instead of generating text, it takes a request plus a fixed set of answer options and returns a calibrated probability for each option, so software can act on the result directly. Microsoft positions it for routing, classification, prioritisation, verification and workflow control inside applications and agents, and offers it through Microsoft Foundry and OpenRouter.

Official source 1 ↗

Features

  • Calibrated probability per option. Given a fixed set of answer options, the model returns a calibrated probability score for each one. The probability is part of the API output, so an application can decide whether to act, defer or ask for review. Official source 1 ↗
  • Yes/no, multiple-choice and rating decisions. Supports yes/no questions, multiple-choice selection and rating scales through a single structured API call. Official source 1 ↗
  • Rubric-based grading. Grades AI responses and agent actions against a rubric, which Microsoft suggests for AI judging and quality control. Official source 1 ↗
  • Single-pass, low-latency scoring. Built by post-training Qwen3.5-9B for fast, single-pass decision scoring. In Microsoft's own benchmarking, its P50 latency was about 35 times faster than GPT-6 Sol. Official source 1 ↗
  • Robustness to rephrasing and reordering. Microsoft reports that across eight kinds of input perturbation the model changed its decision 1.3% of the time on average, with no flips when option descriptions were paraphrased or options were reordered. Official source 1 ↗
  • Safety screening. Microsoft tested the model on 5,250 requests across 11 benchmarks covering harmful content, jailbreaks and prompt injection, and reports that it refused harmful behaviour while keeping utility. Official source 1 ↗

Pricing & free limits

Pay per token

$0.042 per million input tokens; output tokens freeUsage-based, USD per million tokens

Single published rate for API use. Microsoft does not describe tiers, free usage or volume discounts on the launch post. Official source 1 ↗

Billing details

Only input tokens are charged; output tokens are free. Official source 1 ↗

Free access & limits

Free access and its limits were not established by the checked official sources.

Best for & limitations

Agent controls and model routing: deciding whether an agent should continue, stop, retry or hand off, and picking which model should handle a request. Official source 1 ↗

Data labeling: assigning consistent labels to social posts, customer feedback and other data. Microsoft says XBOX Research used it to sort more than 10,000 pieces of player feedback into researcher-defined themes. Official source 1 ↗

AI judging and quality control: scoring chat or agent responses against defined criteria and accepting, revising or rejecting them. Official source 1 ↗

Intent analysis and incident routing: matching a user request to a supported intent, or classifying incidents by type and urgency and routing them to the right team. Official source 1 ↗

Content classification, safety screening and code scanning: categorising content or requests by policy or risk and flagging items for review. Official source 1 ↗

  • Chooses among fixed options; does not generate text. The model scores a fixed set of answer options and returns structured outputs; it is not designed to generate text or reason through open-ended problems like an LLM. Official source 1 ↗
Microsoft-Decision-1 product screenshot
Website previewFixed desktop viewport · 1440 × 900

Similar tools

Published tools that share its categories.

BudaCloud workspace for running a team of AI agents, each with its own computer, coordinated by an Organizer.
Automation & AgentsFree planChecked Oct 11
Visit
Microsoft CopilotMicrosoft's AI assistant built into Word, Excel, PowerPoint, Outlook and Teams, with work-grounded chat, research agents and multi-step task automation.
ProductivityFree planChecked Oct 11
Visit
BusabaseOpen-source database and review workspace where AI agents and people share data, docs and skills.
ProductivityFree planChecked Oct 10
Visit