Microsoft introduced Microsoft-Decision-1 on 9 October 2026, a model for fast decision-scoring that it says is available in Microsoft Foundry and coming soon through OpenRouter. Microsoft prices it at $0.042 per million input tokens, with output tokens free.
Microsoft says the model is designed for routing, classification, prioritization, verification and workflow control, and that it delivers top performance in latency and quality on structured decision tasks, outperforming both LLMs and other decision models.
Business Pill · STRUCTURED OUTPUT
A short explainer of structured output: getting a model to answer in a fixed, machine-readable form. It teaches the general idea only and says nothing about any company in this story.
The key insight: As we read it, Microsoft is pricing a narrow job, picking one option from a fixed list, far below a general model. Its post gives $0.042 per million input tokens with output free, and says the model returns a calibrated probability for each option in a single pass.
According to the post, Microsoft post trained Qwen3.5-9B for fast, single-pass decision scoring and will soon rebase the model on other models, including Microsoft AI (MAI) and OpenAI. Given a fixed set of answer options, it returns a calibrated probability score for each option.
Microsoft says the model supports yes/no, multiple-choice and rating options, as well as rubric-based grading of AI responses and agent actions, through a structured API call.
Relative latency, indexed by us from the multiples in Microsoft’s announcement of 9 October 2026: Microsoft says Microsoft-Decision-1 is 4.5 times quicker than the runner-up, Quyet-1.0-Large, and 35 times quicker than GPT-6 Sol. Microsoft-Decision-1 = 1.
Microsoft’s Benchmarks
Microsoft says Microsoft-Decision-1 achieved the highest accuracy in its 36-benchmark comparison, spanning nearly 150,000 questions across benchmarks kept blind from training. In its benchmarking, the post says, the model was the fastest measured: 4.5 times quicker than Quyet-1.0-Large, the runner-up, and 35 times quicker than GPT-6 Sol.
On robustness, Microsoft says it perturbed the same request in eight ways and that the model changes its decision on 1.3% of perturbations on average. It also says it tested the model on 5,250 safety requests across 11 benchmarks.
The post’s appendix lists the models benchmarked: Quyet-1.0-Large, Surogate Rune 26B-A4B, GPT-6 Luna Decisions, deck-31B, H2O-Lightning-4B and Strands-Decider 2B.
Internal results as reported in Microsoft’s post of 9 October 2026: XBOX Research found the model competitive with GPT-6 Sol, over 14 times faster and 200 times less expensive; Copilot, competitive with GPT5.6 Luna and 100 times faster; Microsoft Discovery, 46 times more consistent at 3 times the speed. These are Microsoft’s own measurements.
How Microsoft Says It Uses It
Microsoft says XBOX Research used the model to sort more than 10,000 pieces of feedback and reviews into fixed themes and found it competitive on quality with GPT-6 Sol while running over 14 times faster and 200 times less expensive.
The post says the Copilot team found it competitive with GPT5.6 Luna and 100 times faster for measuring response quality, and that in Microsoft Discovery it scored as 46 times more consistent than the LLM-based score at three times the speed.
The Structural Read
The model is a small post-trained one. Microsoft says it built Microsoft-Decision-1 by post-training Qwen3.5-9B for single-pass decision scoring, and that it will soon rebase it on other models, including MAI and OpenAI.
The job is the step between agent actions. Microsoft lists agent controls among the uses: evaluating a proposed next step and deciding whether to continue, stop, retry or hand off.
The evidence so far is Microsoft’s own. The 36-benchmark comparison, the latency multiples and the internal results with XBOX Research, Copilot and Microsoft Discovery are all reported in Microsoft’s post.
Microsoft, 9 October 2026
“Now that agentic AI is a reality, we’ve seen that cost plays a major role in how people decide to use AI.”
Three Implications
OUTPUT IS FREE Microsoft says output tokens are free, so the posted price is the $0.042 per million input tokens.
ROUTING BECOMES A MODEL CALL Microsoft says the model is designed for routing, classification, prioritization, verification and workflow control.
DISTRIBUTION BEYOND AZURE Microsoft says the model is coming soon through OpenRouter; the post we read gives no date.
The framework’s starting point: “Inference is a revenue engine: continuous, distributed across millions of applications and agents, incurring marginal cost with every query.”
As we read it, Microsoft-Decision-1 is aimed at that marginal cost: the many small decisions an agent makes between steps, which Microsoft prices per million input tokens with output free.
What Is Not Established
We read Microsoft’s announcement post in full; the benchmark and internal results are Microsoft’s own measurements, not independent tests. The post we read gives no date for OpenRouter availability. We did not contact Microsoft.
Business Engineer Framework
The State of The Inference Economy
A Business Engineer framework on how inference, not training, now decides who makes money in AI.
Microsoft-Decision-1 is a decision-scoring model built by post-training Qwen3.5-9B, available in Microsoft Foundry at $0.042 per million input tokens with free output tokens. Microsoft says it had the highest accuracy in its 36-benchmark comparison and was the fastest model it measured.
94,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. We read Microsoft’s announcement post of 9 October 2026 in full; the benchmark and internal results are Microsoft’s own measurements. We did not contact Microsoft. Nothing here is a forecast, and nothing here is financial or investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
