Microsoft launched Decision-1 on October 9, 2026, a specialized AI model designed for rapid structured decision-making. The model assigns calibrated probability scores to fixed answer options and can process up to 32,000 tokens in a single pass. Microsoft claims it is 4.5 times faster than Quyet-1.0-Large and 35 times faster than GPT-6 Sol, while achieving the highest accuracy across a benchmark of nearly 150,000 questions. Decision-1 is built on Qwen3.5-9B, Alibaba’s open-weight model, and is currently undergoing internal trials for incident response, quality control, and scientific discovery workflows. The model will be available through Microsoft Foundry and OpenRouter.
Source: Read the original article

