Evaluate Microsoft Decision-1 for Lower-Cost AI Routing and Classification Tasks

Microsoft has released Decision-1, a specialized artificial intelligence model designed to separate decision-making from generative tasks. System administrators and developers can use this tool to classify requests, prioritize incidents, and verify AI-generated responses without processing open-ended outputs. The model returns specific probabilities for fixed answer choices, allowing applications to act on the data immediately. This approach differs from standard large language models that produce natural language explanations.

In plain English

General-purpose large language models generate text by predicting the next word in a sequence, which can be slow and expensive for simple yes-or-no questions. Microsoft Decision-1 takes a different approach by accepting a fixed set of possible answers and calculating the likelihood of each one. According to Microsoft, this method reduces both cost and latency compared to using general-purpose models for routine classification tasks. The model handles ratings, rubric-based assessments, and choices among predefined options efficiently.

The background

Microsoft built Decision-1 by post-training the Qwen3.5-9B model and plans to rebase it on other architectures, including Microsoft AI and OpenAI models. The company states that Decision-1 achieved the highest accuracy in a comparison of 36 benchmarks covering nearly 150,000 questions that were kept blind from training. Microsoft also reported that the model operates about 2.5 times faster than the runner-up in those tests. These performance figures are Microsoft’s claims and have not been independently verified by third parties.

What changes now

Developers can access Microsoft-Decision-1 through Microsoft Foundry and OpenRouter. The pricing structure lists the model at 4.2 cents per million input tokens, with no charge for output tokens. In an incident-prioritization example, Microsoft said a request using 86 input tokens cost approximately $0.0000036. At that rate, one dollar would cover more than 275,000 similar decisions. Charlie Dai of Forrester noted that specialized decision models could reduce the cost and latency of agentic AI workflows.

What to do and how to stay safe: Microsoft

• Review current AI workflows to identify tasks that rely on general-purpose models for simple classification or routing decisions.
• Test Decision-1 in a non-production environment to compare its latency and accuracy against existing large language model solutions.
• Monitor the cost per million tokens for your specific use cases, as input token counts can vary significantly by application.
• Check for updates on rebasing plans, as Microsoft intends to rebase the model on Microsoft AI and OpenAI architectures in the future.

General security guidance from the Vector Update newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What is Microsoft Decision-1?

It is a specialized AI model that returns probabilities for predefined answers instead of generating free-form text, designed for tasks like classification and prioritization.

How does it compare to general-purpose LLMs?

Microsoft claims it offers lower costs and latency for decision-making tasks, operating faster and cheaper than general-purpose models for these specific use cases.

Where can developers access the model?

Microsoft-Decision-1 is available through Microsoft Foundry and OpenRouter, with pricing based on input tokens at 4.2 cents per million.