Models
For quick decisions, Microsoft releases an AI that scores options and doesn't write
Microsoft-Decision-1 doesn't write answers: it says which option on a list is the most likely. The figures that put it ahead come from Microsoft itself, and a rival disputes one of them.
The essentials
1 confirmed fact · 2 according to the source · 3 open questions
- Microsoft has released Microsoft-Decision-1, a model for making quick decisions, already available on its Foundry platform.12
- According to the sourceInstead of writing, it gives a probability to each option on a closed list, according to MarkTechPost.2
- According to the sourceMicrosoft says it got 83.5% of the questions right in 36 tests run by the company itself, ahead of its rivals.2
Why it matters
Many tasks don't call for a text, but for a choice: sorting an alert, or deciding whether an assistant carries on or stops. This suggests Microsoft sees a business in small models for that. The catch is that it gives no reasons, so someone has to keep watch.
The details
According to MarkTechPost, Microsoft built it on Qwen3.5-9B, a model from Alibaba, and it only reads text2. It doesn't explain why it picks an option. Its model card asks that it not be left to decide on its own about credit, employment, housing, health or legal rights2. H2O.ai, a rival, disputes the speed comparison: it maintains that the time attributed to it is far higher than the one it measures2.
What we don't know
- Whether the model performs as well in independent tests: all the published figures come from Microsoft.
- Whether it can already be used on OpenRouter: Nadella said it would arrive soon and MarkTechPost treats it as available.
- Who is right about its speed compared with rivals, which H2O.ai disputes.
Sources
- 1Introducing Microsoft-Decision-1, our new model for fast decision-making. It delivers top performance on structured decision tasks, outperforming both LLMs and other decision models in latency and quality. [Read more]
- 2Microsoft AI Releases Microsoft-Decision-1: A Qwen3.5-9B Decision-Scoring Model
Was this story useful? Yes Not really