Microsoft’s Qwen-backed Decision-1 slashes AI decision latency 35x via fixed-set probabilities
Microsoft launched Decision-1, a specialized model that returns calibrated probabilities for predefined answers, aiming to lower costs and latency for structured decision workflows.
Key points
- Microsoft Decision-1 returns probabilities for fixed answer sets rather than free-form text.
- The model is built on post-trained Qwen3.5-9B and available via Microsoft Foundry.
- Microsoft claims the model is 35 times faster than GPT-6 Sol on benchmark tests.
Microsoft has launched Decision-1, an artificial intelligence model designed to return probabilities for predefined answers instead of generating free-form text. This approach allows software applications to consume outputs directly for tasks like classifying requests or prioritizing incidents, according to a company blog post. Microsoft states this method reduces both latency and operational costs compared to using general-purpose large language models for similar structured decisions.
How does Decision-1 differ from standard large language models?
Decision-1 accepts a fixed set of answer options and returns a calibrated probability for each option in a single pass, Microsoft explained. Unlike general-purpose LLMs that produce natural-language explanations, this model provides structured data suitable for immediate programmatic use. Developers can employ these outputs to classify inputs, rank work items, verify responses, or determine the next action in an AI agent workflow without additional parsing.
What performance and cost metrics does Microsoft claim?
Microsoft reported that Decision-1 achieved the highest accuracy across 36 benchmarks covering nearly 150,000 questions that were kept blind from training data. The company also stated the model operates approximately 2.5 times faster than the next-best competitor and 35 times faster than GPT-6 Sol. Pricing is listed at 4.2 cents per million input tokens with no charge for output tokens, meaning $1 could cover more than 275,000 incident-prioritization decisions, according to Microsoft’s calculations.
What is the technical foundation and availability?
The model was created by post-training Qwen3.5-9B, Microsoft said, with plans to rebase it on other architectures including Microsoft AI and OpenAI models in the future. Decision-1 is currently available through Microsoft Foundry and OpenRouter. It supports yes-or-no judgments, choices among predefined options, ratings, and rubric-based assessments of AI responses and agent actions, enabling parallel evaluation of multiple decisions within a single request.
What do industry analysts say about this shift?
Charlie Dai, vice president and principal analyst at Forrester, noted that specialized decision models could reduce the cost and latency associated with agentic AI workflows, according to the report. Microsoft’s benchmark results and cost calculations remain unverified by independent third parties. The separation of decision-making logic from generative text creation suggests a move toward more efficient, specialized AI components for enterprise automation tasks.
Background: Large language models
Large language models vs machine learning differs in scope and output. Machine learning is the broad discipline of algorithms learning from data. Large language models are a specific type of deep learning model focused on understanding and generating human language by predicting the next word in a sequence.
Read the full guide: Large Language Models vs Machine Learning: Core Differences Explained
What to do and how to stay safe: Microsoft
- Review current AI workflows to identify tasks that rely on free-form LLM outputs for simple classification or routing decisions.
- Evaluate whether structured probability outputs from models like Decision-1 could replace complex parsing logic in your applications.
- Monitor vendor benchmarks and pricing models for specialized AI tools to assess potential cost savings against general-purpose models.
- Verify any performance claims independently before integrating new AI components into critical production environments.
Step-by-step guide: Browser Security Settings for Small Business Teams
General security guidance from the TechSec Feed newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Is Microsoft Decision-1 available for public use?
Yes, Microsoft Decision-1 is available through Microsoft Foundry and OpenRouter, according to the company.
What model architecture is Decision-1 based on?
Microsoft built Decision-1 by post-training Qwen3.5-9B, with plans to rebase it on Microsoft AI and OpenAI models.
How much does Microsoft charge for using Decision-1?
Microsoft lists the price at 4.2 cents per million input tokens, with no charge for output tokens.
