Microsoft announced today the Microsoft-Decision-1, a model specialized in structured decision scoring, available now in Microsoft Foundry (and soon on OpenRouter).
Unlike LLMs that generate free text, it receives a fixed set of options and returns a calibrated probability for each in a single pass. It is designed for routing, classification, prioritization, verification, and workflow control in agents and applications.
According to Microsoft, the model (post-trained from Qwen3.5-9B, with plans to rebase on MAI and OpenAI) led in accuracy on a set of 36 benchmarks with nearly 150,000 questions kept blind during training. In latency (P50), it was 4.5× faster than the runner-up (Quyet-1.0-Large) and about 35× faster than GPT-6 Sol.
Pricing is US$0.042 per million input tokens; output is free. Internally, the Xbox Research team already used it to classify more than 10,000 pieces of game feedback (Steam, X, etc.) into defined themes, with quality competitive with GPT-6 Sol, speed >14× higher, and cost roughly 200× lower. Other internal tests (Copilot, incident response, and scientific discovery) also showed gains in speed and consistency.
Satya Nadella highlighted the announcement on X, reinforcing its use in incident response, quality control, and scientific discovery scenarios.
Sources
- Microsoft Command Line — Introducing Microsoft-Decision-1
- Microsoft Community Hub — Introducing Microsoft-Decision-1
- Satya Nadella’s post on X
Transparency: This content was created, edited, or reviewed with the aid of artificial intelligence. The information was cross-checked with public posts on X and sources available on the internet. Consult the original sources for the full context.
By GeekikiBot