Microsoft has launched Decision-1, an AI model built to score, rank and select decisions rather than generate open-ended text, and its first version runs on Qwen3.5-9B, an open-weight model from China's Alibaba Cloud.
The model is available through Microsoft Foundry and is planned for OpenRouter, positioning it against TypeSafe AI's Jev in an emerging category of fast, low-cost "decision models" aimed at AI agents rather than chatbot users.
Microsoft reported 83.5% accuracy across 36 benchmarks covering roughly 150,000 questions, and said Decision-1 runs 2.8 times faster than Jev.
Big savings
The company also claimed Decision-1 costs more than 20 times less than OpenAI's GPT-6 Sol for text-classification tasks, pricing input at $0.042 per million tokens with no charge for output tokens.
Microsoft said it expects to replace or supplement the Qwen foundation with its own models, OpenAI's systems and other technology in a future version.
Still, the build thus far is a little awkward for frontier US technology firms..
Broader push
The move follows Microsoft's push to embed autonomous AI agents across Azure, Microsoft 365 and Dynamics 365 through its MAI Superintelligence team, alongside a parallel effort to bring its cloud-trained MAI model family onto local hardware.
Nvidia, meanwhile, is reportedly in talks to acquire Reflection AI, a US open-model start-up experts believe can rival cheap Chinese alternatives such as DeepSeek.
