Microsoft Releases Decision-1 Model Built on Alibaba's Qwen
Microsoft has released Microsoft-Decision-1, a decision model post-trained on Alibaba's Qwen3.5-9B that it says beats LLMs and Jev 1.13.0 on latency and accuracy in structured decision tasks, and is being tested across all its business areas.
The model was post-trained on Qwen3.5-9B, Microsoft said. The company's Xbox research team has used it to process more than 10,000 pieces of player feedback. Microsoft chief executive Satya Nadella said on social media that, based on the model's performance, the company is testing it across all areas, including incident response, quality control and scientific discovery.
Decision models became a distinct category after the Jev model gained traction in September. Unlike large language models, they do not generate natural language. They output predicted values, probabilities and confidence scores, and they are considerably smaller than LLMs and respond faster. That profile fits AI tasks in which systems must make frequent judgments rather than produce text.
Qwen has become a common base for that work. According to Leiphone, 84 of the 129 open-source decision model projects on HuggingFace are developed on Alibaba's Qwen, a share of more than 65 percent. The list includes decision models released by Microsoft, Perplexity, Cloudflare and H2O.ai.
Microsoft's announcement ties the model's stated gains in latency and accuracy to its use inside Xbox feedback processing, one of the concrete deployments cited alongside the broader testing programme. The areas named by Nadella — incident response, quality control and scientific discovery — are the ones the company said it is currently evaluating, according to the report.