AI news · October 10, 2026
Microsoft releases Decision-1 for fast routing and workflow choices
Microsoft-Decision-1 is available in Foundry and OpenRouter at $0.042/M input with free output; Microsoft reports 35x lower P50 latency than GPT-6 Sol for structured decisions.
- input tokens
- $0.042/M
- reported P50 speed gain
- 35x
- benchmarks compared
- 36
Microsoft introduced Decision-1 on Oct 9 as a small model for fixed-choice scoring rather than open-ended text. The model is post-trained from Qwen3.5-9B, returns a calibrated probability for each option, and supports yes/no, multiple-choice, ratings, classification, rubric grading, routing, and checks on proposed agent actions. It is available through Microsoft Foundry and OpenRouter at $0.042 per million input tokens with free output.
Microsoft reports the highest accuracy in a 36-benchmark comparison covering nearly 150,000 questions, 2.5x the speed of H2O-Lightning-4B, and 35x the speed of GPT-6 Sol. Liquid's open d1 models are the obvious local comparison; Decision-1 is hosted, while d1 targets local and edge devices. These are Microsoft results, so test your labels, thresholds, and false approvals before using the score to trigger an action.
What you can do with it
Try Decision-1 on OpenRouter or Foundry for model routing, support triage, acceptance checks, or agent stop/continue choices. Start with fixed answer options and compare accuracy, latency, and false approvals against your current small model; output is free and input is $0.042/M.
Our take
A cheap, fast decision scorer can remove many small calls from an agent loop, especially when it returns a probability rather than a paragraph. The catch is that Microsoft's comparison is its own benchmark, and a calibrated score is useful only when your answer choices and labels are good.
Links Microsoft announcementTry on OpenRouterTry in Foundry
Source: Microsoft — Made With Models writes the brief; the reporting is theirs.