Microsoft Brings NPU-Optimized DeepSeek-R1 AI Models to Copilot and PCs

Microsoft has announced that it will soon launch “NPU-optimized” versions of the DeepSeek-R1 AI model for Copilot+ PCs, starting with Snapdragon X devices. Support for Intel Lunar Lake and AMD Ryzen AI 9 processors will follow later. The first release, DeepSeek-R1-Distill-Qwen-1.5B, will be available through the Microsoft AI Toolkit, making it easier for developers to create and deploy AI-powered applications directly on their devices. Larger variants, including 7B and 14B models, are expected to arrive later.
According to a Microsoft blog post, these optimized models take full advantage of the powerful NPUs in Copilot+ PCs, enabling efficient on-device AI performance. “With our work on Phi Silica, we were able to harness highly efficient inferencing – delivering very competitive time to first token and throughput rates, while minimally impacting battery life and PC resources,” Microsoft explains. The models also utilize Windows Copilot Runtime (WCR) and ONNX QDQ format for scalability across the Windows ecosystem.
To achieve high performance locally, Microsoft has implemented a sliding window design for super-fast “time to first token” and long-context support, even without dynamic tensor support in the hardware stack. Additionally, the models leverage QuaRot, a 4-bit quantization scheme that maximizes low-bit processing efficiency.
The 1.5B Distilled R1 model will soon be available through the AI Toolkit extension in VS Code. Developers will have access to the Playground to experiment with DeepSeek R1 locally on compatible Copilot+ PCs.
Microsoft is also making DeepSeek R1 models available in the cloud via Azure AI Foundry. This trusted, scalable platform allows businesses to integrate advanced AI seamlessly while meeting SLAs, security, and responsible AI guidelines. Backed by Microsoft’s reliability, the cloud deployment empowers enterprises to adopt AI innovation more broadly.
Controversy Surrounding DeepSeek R1
As Microsoft moves quickly to support DeepSeek R1, the model has sparked controversy in the tech industry. OpenAI has accused DeepSeek of using proprietary code to develop the model, which reportedly cost under $10 million to create—significantly cheaper compared to the billions spent by US firms on their AI models.
Stay tuned for the latest updates as Microsoft continues to push boundaries in AI innovation, despite the challenges and debates surrounding the DeepSeek R1 advancements.
RS News or Research Snipers focuses on technology news with a special focus on mobile technology, tech companies, and the latest trends in the technology industry. RS news has vast experience in covering the latest stories in technology.