Microsoft Launches Decision-1, a Decision Model That Outputs Calibrated Probabilities, Not Text

What happened

Microsoft CEO Satya Nadella announced Microsoft-Decision-1 on X on October 9. It is a decision-only model post-trained on Alibaba's Qwen3.5-9B: given a fixed set of options, it runs a single forward pass and outputs a calibrated probability for each option, without generating text.

The figures Microsoft published are self-reported: first place across 36 benchmarks and nearly 150,000 blind-test questions; P50 latency about 1/35 of GPT-6 Sol and 2.8 times faster than TypeSafe Jev.

Key facts

1. Base: Alibaba Qwen3.5-9B, post-trained
2. Output: calibrated probabilities over fixed options, no text
3. Pricing: $0.042 per million input tokens, output free; 58% cheaper than OpenAI's Decisions API
4. Speed: P50 latency about 1/35 of GPT-6 Sol (company figures)