Microsoft Launches Decision-1 AI Model for Faster, Cheaper Decision-Making

Microsoft has unveiled Microsoft-Decision-1, a new artificial intelligence model built to handle rapid decision-making tasks within software applications and automated workflows. Announced on October 9, 2026, the model is designed to classify information, rank options, assign priority scores, verify conditions and determine which action a system should take next. Unlike conventional large language models that generate text and respond to open-ended questions, Decision-1 focuses on choosing between predefined options and returning structured probability scores. Microsoft says this specialised approach can reduce processing delays and operating costs for developers building AI-powered products. The model is available through Microsoft Foundry, with access through OpenRouter planned for the future.

A key feature of Decision-1 is its emphasis on speed. Microsoft reported that the model achieved the highest accuracy in its comparison of 36 benchmarks covering nearly 150,000 questions. The company also reported that it was 4.5 times faster than Quyet-1.0-Large, the next-fastest model in its comparison, and 35 times faster than GPT-6 Sol in its measured tests. These figures come from Microsoft’s own evaluations and should be interpreted as benchmark results rather than guarantees of performance across every application. Nevertheless, the findings highlight the potential benefits of using a smaller, specialised model for repetitive decisions instead of relying on a larger AI system for every stage of a task.

Decision-1 is built on Qwen3.5-9B, an open-weight model developed by Alibaba, which Microsoft subsequently post-trained for decision-scoring tasks. When given a question and a fixed set of possible answers, it returns a calibrated probability score for each option. Developers can use it for yes-or-no decisions, multiple-choice classification, ratings and rubric-based evaluations of AI-generated responses or proposed agent actions. For example, a customer-service system could use the model to identify the urgency of a complaint, assign it to the appropriate department and determine whether it needs immediate human attention. Microsoft says the model costs $0.042 per million input tokens, while output tokens are free under its announced pricing.

Microsoft is also exploring how Decision-1 can improve its own products and internal operations. The company says its Xbox Research team used the model to categorise more than 10,000 pieces of open-ended feedback from surveys, Steam and social media into predefined themes. According to Microsoft, the system delivered quality comparable to GPT-6 Sol while running more than 14 times faster and costing approximately 200 times less in that particular workload. Other reported applications include evaluating Copilot responses, helping engineers retrieve information during technical incidents and assessing scientific experiments. These examples demonstrate how specialised decision models could become useful components within larger AI systems, particularly where thousands of small decisions must be made quickly and consistently.

The launch reflects a broader shift in AI development towards systems that combine multiple models, each responsible for a different type of task. A general-purpose model may handle complex reasoning or generate a detailed response, while a specialised decision model determines whether the response meets certain criteria, selects the next workflow step or identifies when human review is necessary. Microsoft says Decision-1 was evaluated for robustness and safety, including tests involving harmful requests, jailbreak attempts and prompt injection. However, the model is not designed for open-ended conversation, summarisation or text generation, and Microsoft warns against using it as the sole decision-maker for consequential matters involving employment, credit, housing, healthcare or legal rights. Human oversight and appropriate safeguards remain essential when its scores influence real-world outcomes.

Microsoft-Decision-1 is now available through Microsoft Foundry, giving developers an opportunity to test its capabilities in classification, routing, evaluation and workflow automation. Microsoft has also indicated that future versions may be based on models from its own Microsoft AI portfolio and OpenAI, potentially expanding the range of systems used to power specialised decision-making. Although the company’s benchmark results are promising, developers will need to evaluate the model against their own data, performance requirements and risk levels before adopting it in production. If the approach proves effective at scale, Decision-1 could help reduce the cost of operating AI agents by reserving larger and more expensive models for tasks that genuinely require advanced reasoning, while delegating simpler decisions to faster, more specialised systems.