Microsoft Released Decision-Scoring AI Model
The new model, Microsoft-Decision-1, enables high-speed task routing and classification for developers.
Updated on Oct. 12, 2026 in Artificial Intelligence
Live Poll
Do you trust automated AI tools to accurately score and classify important business decisions?
Microsoft has released Microsoft-Decision-1, a decision-scoring model post-trained from Qwen3.5-9B that is now generally available. The model is designed to provide probabilities for yes/no, multiple-choice, and rating scale inputs through Microsoft Foundry and OpenRouter.
Why it matters
This release provides a specialized, low-cost utility for workflow control and classification tasks, such as automated routing and verification. By separating decision-scoring from generative output, Microsoft aims to streamline complex pipelines that require structured, repeatable logic.
Microsoft-Decision-1 processes data via single-pass scoring and structured API calls. The model demonstrated a 46 times improvement in consistency for experiment evaluation and a 4 times speed increase for adaptive-replanning tasks compared to current standard methods.
The players
Microsoft
A global technology company building enterprise software, cloud infrastructure, and large-scale AI research models.
OpenAI
A research laboratory and developer of the GPT series of large language models, currently in partnership with Microsoft.
The details
The model functions by returning a probability score for each provided answer option rather than generating open-ended text. It was evaluated across 36 benchmarks encompassing 150,000 unique questions. This architecture allows for efficient handling of classification, prioritization, and routing, with feedback processing speeds reported as 14 times faster than GPT-6 Sol when analyzing 10,000 data points.
Timeline
Microsoft-Decision-1 was released as generally available on October 12, 2026.
The Tech Race
This release shifts the focus toward highly efficient, task-specific models that prioritize structured outputs over generative breadth. It challenges existing general-purpose benchmarks by offering a specialized alternative to the GPT-6 Sol and GPT5.6 Luna architectures.
Developers can now integrate the model through Microsoft Foundry and OpenRouter at a cost of $0.042 per million input tokens, with no additional fees for output. The system is suited for workflows requiring high-speed classification and automated decision-making.
The takeaway
The arrival of Microsoft-Decision-1 signals a strategic move toward cost-effective, specialized AI inference in production environments. Watch for future version iterations, as Microsoft has confirmed plans to incorporate base models from their own AI division and OpenAI.
Further reading
For broader trends in enterprise-grade machine learning, explore our Artificial Intelligence section.
Live Poll
Do you trust automated AI tools to accurately score and classify important business decisions?
