Microsoft introduces Decision-1, a model built for fast decisions

Summary

Microsoft has introduced Microsoft-Decision-1, a model that scores a fixed set of answer options and returns calibrated probabilities instead of generating text. Post-trained from Qwen3.5-9B, it targets routing, classification, verification, and agent-control steps, and is available through Microsoft Foundry and OpenRouter. The company reports the highest accuracy across a 36-benchmark comparison and P50 latency about 35 times faster than GPT-6 Sol. All figures are Microsoft's own measurements.

Why it matters for our work

Many steps in agent workflows are quick judgments rather than long generations, and this release treats those judgments as a separate model layer. Matching the model to each step could cut latency and cost in the routine review and routing work shared between humans and AI.

Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.