Microsoft launches Decision-1 AI model for fast decision-making

What is Microsoft-Decision-1

Microsoft has launched Microsoft-Decision-1, a specialized artificial intelligence model built specifically for rapid decision-making and structured software outputs. The system targets low-cost data classification and automated workflow control rather than traditional text generation or open-ended reasoning.

Unlike general-purpose large language models designed to converse with users or draft documents, decision models are engineered to produce instant, machine-readable outputs that software applications can process immediately. Microsoft said the model is intended for task routing, content classification, task prioritization, verification, and workflow management within enterprise applications, autonomous agents, and software pipelines.

The model is now accessible to enterprise developers through Microsoft Foundry, as well as being made available through the third-party OpenRouter platform.

Speed and accuracy benchmark results

In benchmark data released alongside the announcement, Microsoft claimed that Decision-1 outperformed both standard large language models and competing specialized decision models across accuracy metrics and operational latency.

To test the model's performance, Microsoft ran a comparison across 36 distinct benchmarks comprising nearly 150,000 individual questions, with all test datasets kept blind from the model during its training phase. The company reported that Decision-1 achieved the highest overall accuracy score across the complete test suite.

The system recorded even larger performance advantages in execution latency tests. Microsoft reported that Decision-1 was 4.5 times quicker than the second-placed runner-up model, Quyet-1.0-Large, and 35 times faster than general reasoning models such as GPT-6 Sol.

How decision models fit into software workflows

The release highlights a broader shift in artificial intelligence infrastructure towards dedicated, single-purpose models for routine software operations. Large general-purpose language models often incur heavy computational costs and response delays, making them impractical for high-volume background tasks where software simply requires immediate binary or categorised outputs.

Typical use cases for high-speed decision intelligence include determining which downstream AI model should handle an incoming query, analyzing image contents on the fly, or verifying user permissions during automated processing. By focusing strictly on structured scoring and classification at low computational cost, Microsoft aims to enable software developers to build faster, cheaper agentic workflows without relying on slow text generation models.