Microsoft revolutionizes AI with the integration of DeepSeek R1 in Windows Copilot+ PCs

Last update: 30/01/2025

  • Microsoft brings optimized DeepSeek R1 model to Copilot+ PC devices with support for NPUs.
  • The model will enable a more efficient local AI experience with less impact on battery and system resources.
  • More advanced variants of DeepSeek, such as 7B and 14B, will be available at a later stage.
  • Developers will be able to use DeepSeek R1 through Azure AI Foundry and tools like AI Toolkit in Visual Studio Code.
DeepSeek R1 on Windows PCs Copilot+-0

Microsoft has taken a step forward in the use of artificial intelligence by incorporating the DeepSeek R1 model on devices running Windows Copilot+. This advance marks a milestone in the integration of AI solutions, allowing models to be executed locally on hardware optimized with neural processing units (NPUs).

Copilot+ PCs will have access to a “distilled” version of the DeepSeek R1 model, initially in its 1.5B variant, with future updates to include the 7B and 14B models. This will enable more efficient and lightweight AI applications, without depending exclusively on the cloud.

Exclusive content - Click Here  Mothim

A leap in AI efficiency and performance

Optimized DeepSeek R1 models

Microsoft has optimized the DeepSeek R1 model to take full advantage of the performance of the NPUs present in Copilot+ PCs. Thanks to technologies such as Phi Silica and the ONNX QDQ format, the model presents competitive response times, with a speed of up to 16 tokens per second for short inputs. Additionally, the impact on overall system performance and battery consumption is minimized.

One of the keys to this integration is the use of advanced techniques, such as low-precision quantization and a sliding window design. These advances reduce memory usage and optimize the calculation to offer faster answers with compact models, without sacrificing reasoning capabilities.

Ease for developers and companies

Development environments with DeepSeek

DeepSeek R1 integration is not only designed for end users, but also for Empowering the developer experience. Through tools like the AI ​​Toolkit extension for Visual Studio Code, developers can experiment with the model, test it and adapt it to your needs.

Exclusive content - Click Here  Apple says goodbye to the HomePod to focus on the HomePod Mini

The model is also available on Azure AI Foundry, providing companies with a platform reliable and scalable to deploy advanced AI solutions. According to Microsoft, this combination will allow companies to comply with security standards and responsible artificial intelligence regulations.

Initial support and future releases

DeepSeek Technology Innovations

First DeepSeek R1 deployment will target devices with Snapdragon X processors, followed by Intel Lunar Lake and AMD Ryzen AI 9. Microsoft has indicated that this strategy will ensure initial compatibility with the most advanced hardware, ensuring optimal performance.

In the future, more advanced variants are expected to arrive, such as distilled 7B and 14B, which will further expand the possibilities of local artificial intelligence. At the same time, Microsoft will continue to offer the option of cloud-based access through Azure, providing máxima flexibilidad a sus usuarios.

Impacto en el ecosistema tecnológico

Future of DeepSeek

Microsoft's decision to adopt DeepSeek R1 not only highlights the potential of this model, but also its intention to lead the AI ​​market. According to several experts, this step could signal a significant change in the way companies approach the development of artificial intelligence, with an emphasis on Local and autonomous solutions.

Exclusive content - Click Here  How the Citibanamex Pre-Sale Works

Furthermore, the introduction of open source models such as DeepSeek R1 enables a significant reduction in development costs and greater accessibility for startups and independent developers.

With these innovations, Microsoft is facilitating the adoption of artificial intelligence in different sectors, from business applications to creative projects. The expectation is that this technology will allow a transformation in the way we interact with our devices and how we solve complex problems more efficiently and accessibly.