Microsoft Launches Decision Model Built on Alibaba’s Qwen
Microsoft has released Microsoft-Decision-1, a small model that makes decisions between fixed answer options rather than generating text. Built on Alibaba’s open-weights Qwen3.5-9B and post-trained, it offers a unique approach to AI decision-making.
Achint Srivastava, a vice president of software engineering at Microsoft, announced the model in a blog post on October 9th. Microsoft plans to integrate it with other models, including OpenAI’s and their own MAI models, in the near future.
What the Model Does
Microsoft-Decision-1 returns a calibrated probability score for each option given a fixed set of choices, handling tasks like:
- Answering yes or no questions
- Multiple-choice questions
- Ratings
- Evaluating AI responses and agent actions against a rubric
Microsoft suggests potential use cases include routing, classification, prioritization, verification, and workflow control.
Microsoft’s Benchmarks
In internal tests, Microsoft-Decision-1 outperformed its rivals in both accuracy and speed. It was 2.5 times faster than the runner-up and 35 times faster than OpenAI’s GPT-6 Sol. The model changed its decision only 1.3% of the time on average when reworded or reordered requests.
It also passed 11 safety benchmarks covering harmful content, jailbreak attempts, and prompt injection.
Pricing and Availability
Microsoft-Decision-1 costs $0.042 per million input tokens. Input tokens are free for output, available through Microsoft Foundry and OpenRouter, a model router acquired by Stripe in August.
A Growing Category
Microsoft's entry joins a growing field of decision models, with competitors like:
- Jev from TypeSafe AI
- OpenAI’s Decisions API
- Cloudflare, Strands, Liquid AI, Perplexity, and Snowflake, according to The Register.
Over 100 decision models are now available, competing for attention in the rapidly evolving AI landscape.