Microsoft has unveiled Microsoft-Decision-1, a fast decision-scoring AI model, according to a report from Command Line. The model is trained on Qwen3.5-9B, Alibaba's open-weight large language model, and is designed specifically for scoring decisions rather than general conversation.
Microsoft says it will soon rebase the model on MAI, its own in-house model family, as well as on OpenAI and other models. The announcement marks another step in Microsoft's push to build and ship its own AI models alongside its long-standing partnership with OpenAI.
What Microsoft-Decision-1 does
Unlike chatbots built for open-ended dialogue, decision-scoring models evaluate options and assign scores that help systems pick the best path forward. That makes them useful in ranking, recommendation, risk assessment and automated decision pipelines.
The choice of Qwen3.5-9B as the base is notable because it is a relatively compact model, which usually means faster inference and lower cost than much larger systems. Microsoft has not shared benchmark results, pricing or a release date for the model.
- Purpose: fast decision-scoring rather than conversational generation
- Base model: Qwen3.5-9B
- Planned rebase: MAI, OpenAI and other models
- Availability, benchmarks and pricing: not announced
