Microsoft leans on open weight model from Chinese AI lab to challenge Jev

You can now use a Microsoft-branded AI model optimized for structured decisions and cost-efficiency, at speeds claimed to be over twice that of Jev.

Microsoft has launched a new AI model, Microsoft-Decision-1, designed for specific tasks rather than general-purpose text generation. This model is based on Qwen3.5-9B, an open-weight model from Alibaba Cloud, and is available through Microsoft Foundry and OpenRouter. Microsoft claims its decision model is significantly faster and cheaper than competitors like Jev and OpenAI's GPT-6 Sol, particularly for text classification tasks, offering a cost of $0.042 per million input tokens with free output tokens. The company stated that future versions of Microsoft-Decision-1 will utilize homegrown technology. This move positions Microsoft to compete in the growing market for specialized AI decision models, which are seen as crucial for agentic AI applications requiring structured, actionable outputs.

Microsoft's entry into the decision model space, leveraging an open-weight Chinese model, signifies a strategic play to capture a segment of the AI market prioritizing speed, cost, and predictable outputs over general-purpose generation. The rapid proliferation of over 100 such models, including offerings from OpenAI and Cloudflare, indicates a burgeoning demand for AI that performs specific, quantifiable tasks, driven by the increasing viability of agentic AI applications.

This development suggests a market fragmentation where specialized AI models will coexist with broad LLMs, allowing businesses to select the most appropriate tool for specific workflows. Microsoft's dual approach of using an external base model while promising in-house development highlights a competitive strategy focused on both rapid market entry and long-term technological independence in this emerging AI category.