Microsoft launches Decision-1 decision-making model faster and more accurate than Jev

Microsoft unveiled Decision-1 a decision-making model designed to operate faster and with higher accuracy than Jev. The announcement states the model achieves an average accuracy of 83.5% surpassing other decision models including OpenAI GPT-6 and Luna Decision which recorded 79.4%. Microsoft-Decision-1 is adapted from the Qwen3.5-9B model similar to Cloudflare Clef but the company says it will switch to its own Microsoft AI MAI model and OpenAI models in the future.

The training approach focuses on speed and versatile performance tested across multiple benchmarks to prevent narrow expertise. The model has been deployed with the Xbox team where social media content classification achieved quality comparable to larger GPT-6 Sol models while running 14 times faster and at 200 times lower cost.