API
< Back to all clusters
[TECHNOLOGY] · United States · 9 sources

started · updated

Microsoft introduces Decision-1 AI model for structured tasks

Microsoft has introduced Microsoft-Decision-1, a new class of AI model specifically designed for decision-making and classification workloads rather than open-ended text generation. Available through the Microsoft Foundry Models Catalog and OpenRouter, the model is optimized to provide structured outputs, such as JSON, and calibrated probability scores for predefined options.

Unlike traditional large language models (LLMs) that generate conversational responses, Microsoft-Decision-1 focuses on high-volume, low-latency tasks like routing, scoring, and verification. This allows developers to integrate AI into application logic more efficiently by matching specific tasks to specialized models. The model is reportedly based on Alibaba’s Qwen3.5-9B architecture.

In performance benchmarks, Microsoft reported that the model achieved an average accuracy of 83.5% across 36 benchmarks, outperforming competitors such as OpenAI’s GPT-6 Luna Decisions and TypeSafe’s Jev. The model is designed to be significantly faster and more cost-effective for repetitive, structured decision tasks within complex AI agent workflows.

Entities

Alibaba · Microsoft · Microsoft Foundry · OpenAI · OpenRouter · Qwen · Satya Nadella