Microsoft Taps Alibaba Cloud Model for Decision-1 Launch

Microsoft has entered the burgeoning market for decision models with the launch of Microsoft-Decision-1, as first reported by The Register. Unlike standard large language models designed for reasoning or text generation, decision models are built to provide structured outputs for immediate software action.

In a move toward open-weight pragmatism, Microsoft's first version of Decision-1 is based on Qwen3.5-9B, a model developed by Alibaba Cloud, the cloud computing division of Chinese tech giant Alibaba. This mirrors a trend seen with H2O.ai's H2O-Lightning-4B, which is also built on a Qwen model. Microsoft announced the model will be available through Microsoft Foundry and eventually via OpenRouter.

Achint Srivastava, VP of software engineering in the Office of the CTO at Microsoft, stated in a blog post that decision models unlock useful tasks by classifying items and making decisions at high performance and low cost. Microsoft claims Microsoft-Decision-1 is more than 20x cheaper than OpenAI's GPT-6 Sol for text classification, with input tokens priced at $0.042 per million and free output tokens. The company further asserts the model is 2.8x faster than Jev—a model from TypeSafe AI—and 2.5x faster than H2O-Lightning-4B in latency tests, while maintaining 83.5 percent accuracy across 36 benchmarks.

While Microsoft is currently leveraging the Alibaba Cloud model, the company noted it will soon rebase Decision-1 using its own internal models and those from OpenAI, though it did not disclose the reasons for this planned shift.