Microsoft Launches Decision Scoring Model Microsoft-Decision-1

According to a report by Jinse Finance on October 10, Microsoft announced via its official blog the launch of the decision scoring model Microsoft-Decision-1, now available on Microsoft Foundry and OpenRouter. The model is designed for scenarios such as routing, classification, prioritization, validation, and workflow control. Microsoft stated that in blind testing across 36 benchmark comparisons covering nearly 150,000 questions during training, the model achieved the highest accuracy rate and was the fastest in real-world testing: 2.5 times faster than the second-place H2O-Lightning-4B v1.1 and 35 times faster than GPT-6 Sol. The model was trained on Qwen3.5-9B and will later migrate to Microsoft AI (MAI) and OpenAI models. When given fixed options, it outputs calibrated probability scores for each option. In terms of robustness, the average decision reversal rate was only 1.3% across eight types of perturbations for the same request. Pricing is set at $0.042 per million input tokens, with output being free. During internal testing, XBOX Research used the model to process and classify over 10,000 pieces of feedback, achieving quality comparable to GPT-6 Sol but at 14 times the speed and 200 times lower cost.