Microsoft Brings DeepSeek R1 to Windows 11 Copilot+ PCs

Microsoft has announced that it’s bringing NPU-optimized versions of DeepSeek R1 to Copilot+ PCs. The company will also integrate the DeepSeek-R1-Distill-Qwen-1.5B model into its Microsoft AI Toolkit for developers, with the 7B and 14B versions set to follow.

In a recent blog post, Microsoft announced that DeepSeek R1 models will first be available on Qualcomm Snapdragon X-powered PCs, followed by Intel Core Ultra 200V laptops and AMD AI chipsets. This release will allow developers to build AI-powered apps that run locally on compatible Copilot+ PCs.

“The optimized DeepSeek models for the NPU take advantage of several of the key learnings and techniques from that effort, including how we separate out the various parts of the model to drive the best tradeoffs between performance and efficiency, low bit rate quantization and mapping transformers to the NPU,” Microsoft explained.

Microsoft has outlined the hardware requirements for running these AI models on Windows 11 devices. To qualify, a PC must have a Neural Processing Unit (NPU) with at least 40 TOPS (trillion operations per second), 16GB of RAM, and 256GB of storage. This means that PCs with old NPUs won’t be able to run these models locally.

How to run DeepSeek R1 on Windows 11 Copilot+ PCs?

To get started with DeepSeek on a Copilot+ PC, developers will need to create an Azure account on Microsoft’s website. Now, launch Azure AI Foundry and then search for DeepSeek R1. Select the “Check out model” option, click Deploy, and then click “Deploy” again in the pop-up window. The Chat Playground option will appear, and developers can begin experimenting with DeepSeek R1 locally on their Copilot+ PCs.

Microsoft has also announced that it’s making the open-source DeepSeek R1 LLM available for developers through Azure AI Foundry and GitHub. “One of the key advantages of using DeepSeek R1 or any other model on Azure AI Foundry is the speed at which developers can experiment, iterate, and integrate AI into their workflows. With built-in model evaluation tools, they can quickly compare outputs, benchmark performance, and scale AI-powered applications,” said Asha Sharma, Corporate Vice President, AI Platform.

A new report from the Financial Times reveals that Microsoft is investigating whether Chinese startup DeepSeek illegally used OpenAI’s data to train its R1 model. This action violates OpenAI’s terms of service, and Microsoft plans to collaborate with the US government to safeguard its AI model.

Microsoft’s announcement aims to address concerns about DeepSeek potentially storing data on unsecured foreign networks. To mitigate this risk, the company has subjected DeepSeek R1 to rigorous red teaming and safety evaluations to reduce the risk of data breaches.