
Microsoft has launched Microsoft-Decision-1.1, a specialized AI model designed to make fast, low-cost decisions without generating lengthy text responses. Announced on November 17, 2026, the model is now generally available through Azure, joining a growing category of AI systems that prioritize speed, accuracy, and confidence scoring over conversational capabilities.
The announcement coincides with a $3 million pre-seed funding round for British AI startup Sqwish, which is developing a competing decision model called kettei fat. The developments reflect growing demand for specialized AI models that can handle routine decisions more efficiently than conventional large language models.
Decision Models Gain Rapid Adoption
Unlike traditional AI chatbots that generate responses word by word, decision models select from predefined options and assign confidence scores to their choices. These systems are particularly useful for production applications that require rapid classification, routing, or automated decision-making.
In October, decision models accounted for more than 25% of requests processed through Vercel’s AI model router, despite the category emerging only a month earlier. TypeSafe’s Jev model has led early adoption, processing more than 1.5 billion queries through OpenRouter during a recent week.
TypeSafe reached a reported $7.5 billion valuation shortly after introducing Jev, attracting attention from investors and competing AI developers. The model’s early success prompted numerous alternatives, including systems developed by major cloud providers.
Microsoft Introduces Decision-1.1 on Azure
According to Microsoft’s announcement, Decision-1.1 improves upon its earlier model through better accuracy, confidence calibration, and processing speed. Microsoft said the updated system operates more than twice as fast as its predecessor while maintaining low operating costs.
Achint Srivastava, the Microsoft engineering vice president overseeing the project, emphasized that performance on public benchmarks does not always translate into reliable enterprise applications. He said developers should evaluate decision models using real-world tasks that reflect their operational requirements.
Microsoft’s model was developed to balance response time, cost, accuracy, and reliability rather than maximizing benchmark results alone. The company said Decision-1.1 and Jev demonstrate strong performance across these criteria.
Sqwish Raises $3 Million for Customizable Decision Models
Sqwish, founded by Cambridge graduates Ushnish Sengupta and Federica Freddi, is taking a different approach with its kettei fat model. The startup allows customers to continuously improve model performance using their own production data.
Its post-training tools enable businesses to fine-tune decisions for specific workflows without rebuilding their AI systems. Sengupta said this continuous customization is intended to distinguish Sqwish from competitors offering general-purpose decision models.
Freddi expects future AI applications to combine fast decision models with larger reasoning systems, allowing each to handle tasks suited to its capabilities. Sqwish is positioning its technology around that approach as developers increasingly evaluate specialized AI systems for enterprise applications.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.
