Microsoft has introduced a cutting-edge AI model, Microsoft-Decision-1, tailored specifically for structured decision-making tasks. This model has undergone rigorous testing across a variety of scenarios within Microsoft, demonstrating a remarkable speed advantage over existing large language models. In 36 benchmark tests encompassing nearly 150,000 questions, Microsoft-Decision-1 emerged as the leader in accuracy. It excels at selecting the optimal solution from predefined options and provides probability scores for alternative answers. Its capabilities span a wide range of fields, including route optimization and content categorization. Moreover, it can be effortlessly integrated into existing applications, AI agents, and workflow systems.

Microsoft-Decision-1 offers competitive pricing, with input costs as low as $0.042 per million tokens and output being entirely free. Tests reveal that the model performs on par with some large models in relevant tasks, yet it achieves this with significantly faster speeds and lower costs. Built upon the Qwen3.5-9B framework and refined through specialized post-training, Microsoft plans to expand the model's compatibility to other foundational models in the future.

Additionally, Microsoft-Decision-1 showcases exceptional robustness, maintaining an extremely low decision-flipping rate even after perturbation processing and excelling in multiple safety tests. Currently, developers can access the model through Microsoft's AI development platform, with plans to subsequently integrate it into OpenRouter.