DeepSeek R1 7B & 14B models now available for Copilot+ PCs expanding capabilities with Azure AI Foundry

Microsoft has announced the availability of DeepSeek R1 7B and 14B distilled models for Copilot+ PCs through the Azure AI Foundry, marking a significant leap forward in bringing advanced AI capabilities to the edge. This development, revealed on March 3, 2025, builds upon Microsoft’s recent introduction of NPU-optimized versions of the DeepSeek-R1 1.5B distilled model for Copilot+ PCs, further solidifying Windows as the ultimate platform for AI innovation.

DeepSeek R1 7B & 14B models now available for Copilot+ PCs

The introduction of these larger language models to  represents a major milestone in the democratization of artificial intelligence. By enabling 7B and 14B parameter reasoning models to run on Neural Processing Units (NPUs), Microsoft is putting substantial AI power directly into the hands of researchers, developers, and enthusiasts.

These new Windows 11 PCs, powered by Qualcomm Snapdragon X, equipped with NPUs capable of over 40 trillion operations per second (TOPS), are at the forefront of this AI revolution. These purpose-built NPUs offer exceptional efficiency in running AI models locally, balancing speed and power consumption while minimizing impact on battery life, thermal performance, and resource usage. Here are some key features offered:

1. Enhanced reasoning capabilities: The DeepSeek distilled models showcase how even smaller pretrained models can excel with improved reasoning abilities when coupled with NPU hardware.
2. Efficient inferencing: A new scaling law for language models indicates that chain-of-thought reasoning during inference can significantly improve response quality across various tasks.
3. Optimized performance: Microsoft has implemented various techniques, including QuaRot and sliding window for fast first token responses, to optimize the DeepSeek models for efficient operation on Windows 11 PCs.
4. Low-bit inference: Innovations like Phi Silica enable powerful performance with minimal memory and bandwidth requirements, making advanced scenarios like Retrieval Augmented Generation (RAG) and model fine-tuning accessible to application developers.

Availability and compatibility

The rollout of these new AI capabilities begins with Copilot+ PCs powered by Qualcomm Snapdragon X, followed by support for Intel Core Ultra 200V and AMD Ryzen processors.

Developer access and tools

Developers can easily access all distilled variants (1.5B, 7B, and 14B) of DeepSeek models through the AI Toolkit VS Code extension (See AI Toolkit for Visual Studio Code for more information). The Azure AI Foundry provides optimized ONNX QDQ format models, allowing for seamless local deployment and experimentation.

Cloud and Edge Synergy

Microsoft’s vision extends beyond local processing, offering a synergy between Copilot+ PCs and Azure services. This approach enables developers to leverage both local compute capabilities for smaller tasks and cloud resources for larger, more intensive workloads.

Similar Posts

1. Microsoft’s innovative Azure AI features unveiled aiming to enhance AI chatbot security
2. Microsoft Visual Studio Professional 2022 limited time offer: A developer’s dream at a fraction of the cost
3. Microsoft Teams supercharges collaboration with better, AI-powered Copilot features at your disposal
4. Microsoft Introduces new Surface Pro 10 and Surface Laptop 6 at “New Era of Work” livestream
5. Microsoft’s Surface event: How to watch and 5 exciting things to expect from Copilot and Windows happening today at noon