Microsoft launches Decision-1 model, 35x faster than GPT-6 Sol

Microsoft has launched Microsoft-Decision-1, a decision model built for structured decision tasks. The company says it delivers top-level performance in both response speed and decision quality, outperforming large language models and traditional decision models. It is now live on Microsoft Foundry and OpenRouter.

In a comparison covering 36 benchmarks and nearly 150,000 blind test items, the model scored the highest accuracy. It also ranked first in response speed, running 4.5 times faster than runner-up Quyet-1.0-Large and 35 times faster than GPT-6 Sol.

Microsoft built the model by post-training Qwen3.5-9B, enabling single-pass decision scoring with very low latency, and plans to adapt it to other base models like MAI and OpenAI. Safety tests across 11 benchmarks and 5,250 requests show it can block harmful content, jailbreaks, and prompt injections while staying highly usable.

On pricing, it costs $0.042 per million input tokens, while output is free.