Microsoft just announced that it will be releasing NPU-optimized versions of DeepSeek-R1! This will take advantage of the AI-optimized hardware that Copilot+ PCs have. According to the Windows Blog, this feature will first come to PCs with the Qualcomm Snapdragon X, and then to those with Intel Core Ultra 200V (Lunar Lake) and other processors. The initial launch will feature the DeepSeek-R1-Distill-Qwen-1.5B model, which a UC Berkeley AI research team claims is the smallest that produces correct answers. However, Larger models with 7 and 14 billion of parameters will arrive shortly after. 🤖

DeepSeek's optimizations allowed it to require 11 times less processing power compared to its Western competitors, making it a ideal model to run on devices of consumption. In addition, it uses the Windows Copilot Runtime, allowing developers to use DeepSeek APIs in their applications. 💻

In addition, Microsoft claims that this optimized version of DeepSeek NPU will offer “very competitive response times and throughput rates, minimizing the impact on battery life and resource consumption on PCs.” This means that Copilot+ users will be able to use the latest version of DeepSeek NPU. You can expect the power and performance of models competitors like Meta's Llama 3 and OpenAI's o1, without sacrificing battery life. 🔋

DeepSeek Copilot+: The Revolution in PCs for Programmers.

However, the availability of DeepSeek in the PCs Copilot+ is more focused on programmers and developers than consumers. Perhaps Microsoft is using it to encourage them to create more applications that take advantage of the PCs AI, as many people still don't see the need for them, and market research suggests that users only buy these devices because they are the only option available today. 📊

Another thing that we found intriguing is the preferential treatment of Microsoft towards Qualcomm Snapdragon X-powered PCs right now. While they launched the Copilot+ branding with these chips last July, the latest laptops from Intel and AMD AMD also features built-in NPUs. AMD has even published instructions on how users can run DeepSeek on NPUs. Ryzen AI and Radeon GPUs, and the company claims the RTX 7900 XTX runs DeepSeek better than the RTX 4090. 🖥️

Regardless, we remain excited about the possibilities that DeepSeek brings to the table. artificial intelligence. Being from open source, almost anyone can download and run it locally, allowing others to build on the advancements and optimizations the original model has made. 🚀