The model is designed to score a fixed set of options and return calibrated probabilities through a structured call, rather than generate open-ended text. The report says it supports yes/no, multiple-choice, rating, and rubric-based scoring.
Microsoft says the model was post-trained on Qwen3.5-9B and topped its 36-benchmark comparison, covering nearly 150,000 questions. The company also reports P50 latency 4.5 times faster than Quyet-1.0-Large and 35 times faster than GPT-6 Sol. These are company-reported results, not independent verification. The report gives starting input pricing of $0.042 per million tokens, with output tokens free, and says OpenRouter support is planned. Source: https://www.testingcatalog.com/microsoft-launches-decision-1-model-in-foundry