--- title: Microsoft separates decision-making from generative AI... description: Microsoft has launched Decision-1, a specialized AI model that returns calibrated probabilities for predefined answer sets instead of generating free-form... canonical: https://daily.dev/posts/microsoft-separates-decision-making-from-generative-ai-with-a-new-model-z7lrvkhp8 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Microsoft separates decision-making from generative AI with a new model | daily.dev og:description: Microsoft has launched Decision-1, a specialized AI model that returns calibrated probabilities for predefined answer sets instead of generating free-form... og:url: https://daily.dev/posts/microsoft-separates-decision-making-from-generative-ai-with-a-new-model-z7lrvkhp8 og:image: https://api.daily.dev/og/posts/Z7LrVkHp8.png og:image:alt: Microsoft separates decision-making from generative AI with a new model og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Microsoft separates decision-making from generative AI with a new model **[InfoWorld](https://daily.dev/sources/infoworld)** · 4 min read · 0 upvotes · 0 comments ## Summary Microsoft has launched Decision-1, a specialized AI model that returns calibrated probabilities for predefined answer sets instead of generating free-form text, aimed at routing, classification, incident prioritization, and verifying AI-generated responses. Built by post-training Qwen3.5-9B, it is available via Microsoft Foundry and OpenRouter, with plans to rebase it on other models including MAI and OpenAI. Microsoft claims top accuracy across 36 benchmarks and far lower cost and latency than general-purpose LLMs like GPT-6 Sol, pricing input tokens at 4.2 cents per million with no output charge. Analysts note the approach could cut costs in agentic workflows but warn that Decision-1 offers no rationale for its scores, that wording changes can shift results, and that enterprises should test calibration, log decisions, and require human review for high-impact cases rather than trust vendor benchmarks directly. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is Microsoft's Decision-1 model and what is it used for? Decision-1 is an AI model from Microsoft that returns calibrated probabilities for a fixed set of predefined answers in a single pass, instead of generating free-form text like a general-purpose LLM. It is designed for tasks such as classifying requests, prioritizing incidents, verifying AI-generated responses, and determining next actions in agent workflows. It was built by post-training Qwen3.5-9B and is available through Microsoft Foundry and OpenRouter. _Teams designing agentic AI pipelines can follow daily.dev for updates on specialized decision models versus general LLMs._ ### How much does Microsoft Decision-1 cost compared to using a general-purpose LLM for routing tasks? Microsoft prices Decision-1 at 4.2 cents per million input tokens with no charge for output tokens. In an incident-prioritization example using 86 input tokens and two output tokens, the cost was approximately $0.0000036, meaning $1 would cover more than 275,000 similar decisions. Microsoft also claims it is about 2.5 times faster than the runner-up model and 35 times faster than GPT-6 Sol, though these figures are the vendor's own unverified claims. _Engineers weighing cost-per-decision tradeoffs in AI agent architecture can track this comparison on daily.dev._ ### What are the governance risks of using a probability-based decision model like Decision-1 in production? Decision-1 returns scores without any rationale, making decisions harder to explain, validate, or challenge, according to analysts. Results can also shift depending on how questions are worded. Recommended mitigations include calibration testing, controls over question wording and score thresholds, audit logging, and mandatory human review for high-impact decisions, especially where customer outcomes or regulatory obligations are at stake. _Architects adding guardrails to agentic AI workflows can follow ongoing coverage of this tradeoff on daily.dev._ ## Similar posts on daily.dev - [Microsoft just admitted its biggest AI mistake — and spent $2.5 billion fixing it](https://daily.dev/posts/microsoft-just-admitted-its-biggest-ai-mistake-and-spent-2-5-billion-fixing-it-cxuqalrpi) · The New Stack · 0 upvotes · 0 comments - [Microsoft joins AI cost-cutting trend by relying more on its own models](https://daily.dev/posts/microsoft-joins-ai-cost-cutting-trend-by-relying-more-on-its-own-models-paf5bf3lb) · TechCrunch · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#microsoft](https://daily.dev/tags/microsoft), [#agentic-ai](https://daily.dev/tags/agentic-ai), [#ai-governance](https://daily.dev/tags/ai-governance) [View this post on daily.dev](https://daily.dev/posts/microsoft-separates-decision-making-from-generative-ai-with-a-new-model-z7lrvkhp8) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Microsoft separates decision-making from generative AI with a new model","url":"https://daily.dev/posts/microsoft-separates-decision-making-from-generative-ai-with-a-new-model-z7lrvkhp8","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/microsoft-separates-decision-making-from-generative-ai-with-a-new-model-z7lrvkhp8"},"datePublished":"2026-10-13T11:34:23.176Z","dateModified":"2026-10-13T13:40:55.077Z","description":"Microsoft has launched Decision-1, a specialized AI model that returns calibrated probabilities for predefined answer sets instead of generating free-form...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/74716998ef3baa4eaa46e74beb989eb1?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/74716998ef3baa4eaa46e74beb989eb1?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"InfoWorld","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"InfoWorld","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/bf6d68a999064029b0bb09aa6268f1f3","url":"https://daily.dev/sources/infoworld"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/microsoft-separates-decision-making-from-generative-ai-with-a-new-model-z7lrvkhp8","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,microsoft,agentic-ai,ai-governance","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"InfoWorld","item":"https://daily.dev/sources/infoworld"},{"@type":"ListItem","position":3,"name":"Microsoft separates decision-making from generative AI with a new model"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/microsoft-separates-decision-making-from-generative-ai-with-a-new-model-z7lrvkhp8#faq","mainEntity":[{"@type":"Question","name":"What is Microsoft's Decision-1 model and what is it used for?","acceptedAnswer":{"@type":"Answer","text":"Decision-1 is an AI model from Microsoft that returns calibrated probabilities for a fixed set of predefined answers in a single pass, instead of generating free-form text like a general-purpose LLM. It is designed for tasks such as classifying requests, prioritizing incidents, verifying AI-generated responses, and determining next actions in agent workflows. It was built by post-training Qwen3.5-9B and is available through Microsoft Foundry and OpenRouter. Teams designing agentic AI pipelines can follow daily.dev for updates on specialized decision models versus general LLMs."}},{"@type":"Question","name":"How much does Microsoft Decision-1 cost compared to using a general-purpose LLM for routing tasks?","acceptedAnswer":{"@type":"Answer","text":"Microsoft prices Decision-1 at 4.2 cents per million input tokens with no charge for output tokens. In an incident-prioritization example using 86 input tokens and two output tokens, the cost was approximately $0.0000036, meaning $1 would cover more than 275,000 similar decisions. Microsoft also claims it is about 2.5 times faster than the runner-up model and 35 times faster than GPT-6 Sol, though these figures are the vendor's own unverified claims. Engineers weighing cost-per-decision tradeoffs in AI agent architecture can track this comparison on daily.dev."}},{"@type":"Question","name":"What are the governance risks of using a probability-based decision model like Decision-1 in production?","acceptedAnswer":{"@type":"Answer","text":"Decision-1 returns scores without any rationale, making decisions harder to explain, validate, or challenge, according to analysts. Results can also shift depending on how questions are worded. Recommended mitigations include calibration testing, controls over question wording and score thresholds, audit logging, and mandatory human review for high-impact decisions, especially where customer outcomes or regulatory obligations are at stake. Architects adding guardrails to agentic AI workflows can follow ongoing coverage of this tradeoff on daily.dev."}}]} ```