Microsoft Unveils Decision-1, an AI-Based Decision-Making Model from Alibaba

The model is designed for tasks such as classifying information, prioritizing requests, reviewing AI responses, and managing workflows.

This model is used for tasks such as categorizing information, prioritizing requests, checking AI responses, and controlling workflow.

Become a special member of Arjan Ambassador in order to benefit from and access special facilities and different departments in Arjan Ambassador.

Microsoft has introduced the Microsoft Decision-1 artificial intelligence model, which instead of generating text, evaluates the available options and provides a probability score for each option. The model is based on Qwen3.5-9B, Alibaba's heavyweight model, and Microsoft says it outperformed competing decision models in terms of accuracy and speed in its internal tests.

According to Microsoft, this model is used for tasks such as categorizing information, prioritizing requests, reviewing AI responses, and controlling workflows. Its input processing fee is $0.042 per million tokens and output tokens are free.

Microsoft developed Microsoft-Decision-1 with post-training model Qwen3.5-9B belonging to Alibaba. According to this company, future versions of the model will be developed based on other models, including Microsoft's proprietary models called MAI and OpenAI models. Microsoft has not announced a reason for this change in its article.

The Microsoft-Decision-1 model is designed to choose among predefined options, unlike large language models that generate text. The model provides a probability score for each option and supports yes or no questions, multiple choice questions, and scoring. It can also evaluate the responses of AI models and the actions of AI agents based on specific criteria.

This model is intended for applications such as selecting the appropriate model to process each request, categorizing information, prioritizing tasks, validating data, and controlling workflow. The model can also examine the responses and proposed actions of artificial intelligence agents and help to decide whether to continue working, repeat the operation or refer it to other tools or models.

Microsoft-Decision-1 accepts only text input and has a context window of 32,000 tokens. Microsoft also cautions against using this model as the sole basis for important decisions about people, such as hiring or granting credit.

Microsoft says that Microsoft-Decision-1 achieved the highest accuracy in a comparison of 36 benchmarks. This evaluation included nearly 150,000 questions that were not provided to the model at the time of training.

According to the company's tests, the new model is 2.5 times faster than H2O-Lightning-4B v1.1, the second-ranked competitor in this comparison. Microsoft also says it was 35 times faster than GPT-6 Sol in its tests.

This model is now available to developers through the Microsoft Foundry platform and the OpenRouter service. Its entry fee is $0.042 per million tokens and there is no cost to process outgoing tokens.

I am a linguistics student and as much as I am interested in "words", I also enjoy exploring the world of technology and science.