Microsoft has introduced Microsoft-Decision-1, a specialised artificial intelligence model designed to handle structured decision-making tasks at high speed and lower cost. The model aims to improve the efficiency of AI agents and business applications by handling routine decisions without repeatedly relying on larger, more expensive language models.
According to Microsoft's published findings, Decision-1's median-latency performance is approximately 35 times faster than GPT-6 Sol and 4.5 times faster than Quyet-1.0-Large in the company's reported comparisons.
ALSO READ: Anthropic Cites New AI Misbehaviour, Including On US Government Websites
Built on the Qwen3.5-9B model, Decision-1 has undergone specialised post-training to improve decision scoring. Unlike general-purpose AI systems that focus on generating text and answering complex questions, it is designed to select from predefined options and assign probability scores to possible outcomes.
Microsoft Reports Benchmark and Cost Advantages
Microsoft said Decision-1 ranked first in accuracy across 36 benchmark tests covering nearly 150,000 questions. The results were reported by the company and have not been independently verified.
The model is priced at $0.042 per million input tokens, while output tokens are free, according to the published pricing details. The pricing structure is intended to make high-volume, repetitive decision-making tasks more economical for businesses and developers.
Microsoft's Xbox Research team also tested the model on more than 10,000 pieces of game feedback and reviews. According to the company, Decision-1 delivered comparable quality to GPT-6 Sol while operating more than 14 times faster and at approximately one-two-hundredth of the cost.
Designed for AI Agents and Business Workflows
Also Read | Layoffs Are Rising In Tech, But These Skills Are Helping People Unlock Better Pay
Decision-1 supports yes-or-no decisions, multiple-choice selections, ratings and rubric-based evaluations. Its potential applications include content classification, application routing, task prioritisation, quality control, result verification and safety screening.
Within AI-agent workflows, the model can help determine whether a task should continue, an operation should be retried, an issue should be escalated or human review is required. These capabilities could reduce processing time and operating costs in applications handling large volumes of decisions.
Security Testing and Availability
Microsoft reported testing Decision-1 for harmful-content identification, jailbreak attempts and prompt injection. The company also measured its consistency under different input variations, reporting an average decision-change probability of 1.3% across eight perturbation methods.
These findings are based on Microsoft's evaluations and require independent validation. Decision-1 is available to developers through Microsoft Foundry, while OpenRouter access has been planned.
ALSO READ: AI Trade's Weak Link Isn't Chips, It's Debt: Who Is Most Exposed If Yields Keep Climbing
The launch reflects Microsoft's efforts to make AI systems more efficient by separating routine decision-making from more demanding reasoning and text-generation tasks. However, independent evaluations will be important in determining whether the reported speed, accuracy and cost advantages hold across real-world applications.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
