Microsoft launches Decision-1, a 9B Qwen decision scorer
Original title: Microsoft Decision-1: 9B Qwen Model Launch Explained
The model returns probabilities over a fixed option list and is on Foundry now, with OpenRouter access following.
On October 9, 2026, Microsoft introduced Microsoft-Decision-1, a model that scores predefined answer options in one forward pass instead of generating text. It is post-trained from Alibaba’s Qwen3.5-9B and offered first through Microsoft Foundry, then OpenRouter. Microsoft cites routing, classification, prioritization, verification, and workflow control, and says it will later rebase the model on Microsoft AI and OpenAI foundations. Microsoft claims 83.5 percent accuracy across 36 benchmarks and about 150,000 held-out questions, plus large speed gains versus named comparison models.
Unconfirmed: Accuracy and speed figures are Microsoft’s claims, also relayed by other outlets, not independently verified here.
- Input is priced at $0.042 per million tokens, with output listed as free.
- Microsoft claims 4.5 times the speed of Quyet-1.0-Large and about 35 times GPT-6 Sol.
- Named uses include incident response and scientific discovery, according to the company.
- The Qwen base is described as temporary scaffolding, not the long-term foundation.
- Products
- Microsoft FoundryOpenRouter
- Organizations
- OpenAI
- AI models
- Microsoft-Decision-1Qwen3.5-9B
From the original article
13 min read at shattered.io
Microsoft used October 9, 2026 to introduce a model that refuses to write you a paragraph. Microsoft-Decision-1 doesn’t chat, summarize, or draft emails. It scores a fixed list of answer options in a single forward pass and hands back a probability for each one, then gets out of the way.
By gmays · shattered.io
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.