Microsoft skipped OpenAI's decision model and built its own on Alibaba's Qwen

Microsoft builds its own decision model

Microsoft has skipped OpenAI's decision model and developed its own, built on Alibaba's Qwen, according to a report from The New Stack. The move suggests that even close AI partners are hedging their bets when it comes to the next generation of AI infrastructure.

Decision models, which determine what an AI system should do next, have become one of the hottest segments in the industry. OpenAI recently made its decision model public, and a separate effort involving AWS, Upstage and Ollama to agree on a common decision-model API has drawn attention, though OpenAI has not signed on to that initiative.

What is known so far

The report does not include details on when the Microsoft-built model will ship, its performance figures, or which products it will power. What is clear is that basing the model on Qwen, Alibaba's model family, gives Microsoft a degree of independence from OpenAI's technology roadmap at a time when decision models are suddenly everywhere.

Industry context

The news lands amid rapid change across the AI stack. GitHub Copilot is going local, though Microsoft has not said what data gets sent to the cloud. Anthropic's Claude Sonnet 5.5 has been positioned as significantly cheaper than Opus 5.5. And as decision models multiply, AWS, Upstage and Ollama are already agreeing on shared APIs while OpenAI stays on the sidelines.

For developers and enterprises, the growing number of model choices also raises familiar questions about portability, cost and vendor lock-in as decision models become a distinct layer of the AI stack.

Sources: The New Stack