Microsoft launches Decision-1: an AI model that decides instead of writing
The small model assigns probabilities to fixed answer options and costs next to nothing. OpenAI is also betting on this new class of model with its Decisions API.
Photo: Brian Smale and Microsoft / Wikimedia Commons, CC BY-SA 4.0 (cropped and resized)
Video: YouTube · OpenAI · youtube.com
Microsoft on 9 October unveiled "Microsoft-Decision-1", a model for fast decision-making. Instead of generating free text, it receives a fixed set of options – such as yes/no, multiple choice or ratings – and returns a calibrated probability for each. It is designed for tasks such as routing requests, classification, checking results and controlling AI agents. It is built on Alibaba's open model Qwen3.5-9B, which Microsoft post-trained.
According to Microsoft's own tests, Decision-1 achieved the highest accuracy in a 36-benchmark comparison with nearly 150,000 questions and was 35 times faster than OpenAI's GPT-6 Sol. At Xbox Research it labelled more than 10,000 pieces of player feedback – with quality comparable to GPT-6 Sol, Microsoft says, but more than 14 times faster and 200 times cheaper. The model is available in the Microsoft Foundry cloud service and costs $0.042 per million input tokens; output is free.
A new class of model is taking shape: since 6 October OpenAI has offered its Decisions API, based on GPT-6 Luna, to all developers as a public beta. It is meant to answer up to ten times faster than GPT-6 Luna through the standard Responses API and costs $0.10 per million input tokens, also with no charge for output.