Bing and Copilot Test Microsoft-Decision-1 for Query Routing
Microsoft is testing Microsoft-Decision-1, a decision-scoring model it introduced last week, inside both Bing Search and Copilot. The model is built for routing, classification, prioritisation, verification and workflow control, and Microsoft says it beats both large language models and other decision models on structured decision tasks while keeping latency low.
Rangan Majumder, a Microsoft engineer, said on X that the company is experimenting with the model in Copilot and that it is "pretty good", adding he was impressed by how quickly it improved. When asked whether Bing is using it, Krishna Madhavan of the Bing team replied that the team is "always experimenting". The model is available through Microsoft Foundry and OpenRouter.
The testing does not mean every Bing query or every user is routed through the model. For brands tracking how they appear in AI answers, the detail matters because routing and classification models decide which queries get an AI-generated response at all, and which sources are pulled into it. A change in that layer can shift where citations land without any change to the underlying content.
Reported by Search Engine Roundtable ↗