Microsoft launches Decision-1 model in Foundry
Microsoft is pitching Decision-1 as a cheap, fast scoring layer for software that needs to choose, not chat.
The model takes fixed options and returns calibrated probabilities through a structured API, covering yes/no, multiple choice, ratings, and rubric grading. Microsoft says it beat other tested models on latency and quality across a withheld benchmark set, and cites internal uses in Xbox feedback labeling, Copilot workflows, incident retrieval, and scientific replanning. It is available in Microsoft Foundry now, with OpenRouter support planned. Pricing starts at $0.042 per million input tokens, with output tokens free. TestingCatalog's note
The model takes fixed options and returns calibrated probabilities through a structured API, covering yes/no, multiple choice, ratings, and rubric grading. Microsoft says it beat other tested models on latency and quality across a withheld benchmark set, and cites internal uses in Xbox feedback labeling, Copilot workflows, incident retrieval, and scientific replanning. It is available in Microsoft Foundry now, with OpenRouter support planned. Pricing starts at $0.042 per million input tokens, with output tokens free. TestingCatalog's note
score 5