Microsoft unveils Microsoft-Decision-1, a fast decision-scoring model trained on Qwen3.5-9B, and says it will soon rebase it on MAI, OpenAI, and other models

Your AI agents can now make decisions and classify information 2.5x faster and with higher accuracy than before.

Microsoft has introduced Microsoft-Decision-1, a new model designed for rapid decision-scoring tasks. Available through Microsoft Foundry and OpenRouter, this model specializes in functions like routing, classification, prioritization, verification, and workflow control, aiming to provide structured outputs for immediate software action. The company claims Decision-1 outperforms both large language models (LLMs) and other decision models in terms of latency and accuracy. In internal benchmarks, it achieved the highest accuracy across 36 tests involving approximately 150,000 questions and was significantly faster, reportedly 2.5 times quicker than its closest competitor, H2O-Lightning-4B v1.1, and 35 times faster than GPT-6 Sol. Microsoft plans to rebase Decision-1 on models like MAI and OpenAI's offerings in the future.

Decision models represent a distinct category of AI, separate from generative LLMs, optimized for structured, actionable outputs. Microsoft's entry with Decision-1 signals a growing market for AI systems that can perform specific tasks like routing and prioritization with high efficiency and low cost. This focus on specialized, performant models could lead to more integrated and cost-effective AI deployments across various software applications and agentic workflows.

The model's performance claims, particularly its speed advantage over existing LLMs, suggest a potential shift in how companies approach AI integration for tasks requiring rapid, reliable decision-making. As Microsoft plans to leverage other advanced models like MAI and OpenAI's technologies for future iterations, Decision-1's capabilities may further expand, influencing the development of more sophisticated AI-driven automation and control systems.