Microsoft has recently officially launched a new generation of AI model focused on structured decision-making tasks - Microsoft-Decision-1. The model has already completed field tests in multiple internal scenarios, demonstrating significant advantages in terms of speed. In 36 authoritative benchmark tests covering approximately 150,000 questions, its accuracy consistently ranked first, making it a highly anticipated high-speed decision-making star.

In terms of core functions and application scenarios, this model is mainly used to accurately select the optimal solution from predefined options and can assign corresponding probability scores to each alternative answer. Its functionality can seamlessly cover practical business scenarios such as routing distribution and content classification, and can be easily integrated into existing applications, AI agents, and workflow systems, significantly improving the efficiency of complex workflows.

In terms of technical architecture and foundation models, Microsoft-Decision-1 is built upon Qwen3.5-9B and has undergone specialized post-training. Microsoft has also revealed that it plans to migrate the model to more other base models in the future. In terms of stability and security, the model demonstrates excellent robustness, with a very low decision reversal rate after undergoing various perturbation treatments, and has also achieved good results in multiple strict security tests.

Additionally, the model is highly competitive in terms of economic cost and commercial pricing. Its pricing standard is only $0.042 per million input tokens, while output is completely free. Actual testing shows that it can maintain a comparable quality level with some large models in similar decision-making tasks, but achieves a breakthrough in processing speed and significantly reduces usage costs.

Currently, the model has been officially made available to developers through Microsoft's AI development platform. Furthermore, Microsoft plans to integrate it into the OpenRouter platform in the future, providing developers with more convenient access support.