Microsoft launches Decision-1 for structured decisions

On October 10, 2026, Microsoft CEO Satya Nadella announced the launch of Microsoft-Decision-1, a model built for structured decision-making tasks. According to a Wall Street Journal-affiliated report published that day, the model picks the best option from a preset set of choices and assigns a probability score to each candidate answer, helping downstream applications decide whether to proceed, retry, escalate, or hand the case to a human reviewer. Nadella said in a post on X that the model performs well on structured decision tasks and beats large language models and other decision models on both latency and quality.

Microsoft's stated performance and pricing figures are as follows: in internal testing, Microsoft-Decision-1 runs 35 times faster than OpenAI's GPT-6 Sol and 4.5 times faster than Quyet-1.0-Large. On cost, input is priced at $0.042 per million tokens, while output is free — a structure the report describes as attractive for high-frequency, repetitive decision applications. Microsoft also said the model has been tested internally for incident response, quality control, and scientific research, and that it posted results across 36 benchmarks covering nearly 150,000 questions, though the report's summary cuts off before giving specific scores.

What can be verified so far stops at the launch and internal testing stage: the model has been released by Microsoft, and the performance comparisons and pricing come from Microsoft alone. The report does not mention third-party reproduction or external evaluation results.