The news surrounding Microsoft Copilot+ highlights several advancements and challenges associated with the integration of the DeepSeek R1 AI models into Windows 11 PCs.
Battery Life is notably improved, particularly with the Asus Zenbook S 14, which boasts double the battery life compared to previous models. This enhancement is attributed to the efficient inferencing capabilities of the NPU-optimized DeepSeek R1 models, which are designed to minimize the impact on battery life while executing AI applications. This efficiency is crucial for users who rely on portable devices for extended periods.
In terms of System Performance, the introduction of NPU-optimized versions of the DeepSeek R1 models is expected to significantly enhance AI application performance on-device. The models support various processor architectures, which allows for a broader range of devices to benefit from improved performance. However, there are concerns regarding older NPU-equipped PCs that may not support local model execution, potentially limiting performance gains for some users.
When discussing Hardware, the Asus Zenbook S 14 showcases a stylish and ultra-light design, constructed from environmentally friendly materials. It meets military standards for durability, which is a positive aspect for users concerned about build quality. However, the device's glossy display may lead to reflections in bright environments, which could detract from the user experience.
In the realm of Software, the integration of the DeepSeek R1 models into Copilot+ PCs allows for efficient on-device AI application development. The availability of an AI Toolkit for developers to experiment with these models locally is a significant advantage. However, the rapid development pace may lead to stability or support issues, and there are concerns about the complexity of implementing new features, particularly for developers unfamiliar with the AI Toolkit.
Security and Privacy are enhanced by the ability to interact with advanced AI models locally, which reduces the need for cloud processing and can mitigate potential data exposure risks. This local execution is a key selling point for users who prioritize privacy in their computing environments.
Compatibility remains a concern, as there may be issues with non-optimized hardware and potential limitations in supporting dynamic input shapes and behaviors in certain model variants. This could restrict the user base and create challenges for developers aiming to implement these models across diverse hardware configurations.
Lastly, Reliability is a critical factor, especially given the potential learning curve for developers adapting to the new AI Toolkit and the complexities associated with the ONNX QDQ format. The limited initial availability of models may also affect the overall reliability of the platform as it scales.
Overall, while Microsoft Copilot+ and the DeepSeek R1 models present exciting advancements in AI capabilities for Windows 11 PCs, they also introduce challenges that users and developers will need to navigate as the technology evolves.
This is not a Copilot+ review.
Source: https://wccftech.com/deekseek-r1-npu-optimized-version-to-run-on-copilot-windows-11-pcs/
| Key positive aspects of Copilot+ | Key concerns related to Copilot+ |
|---|---|
| NPU-optimized DeepSeek R1 models enhance AI application performance on-device. | Older NPU-equipped PCs may not support local model execution. |
| Efficient inferencing with minimal impact on battery life and PC resources. | Potential compatibility issues with non-optimized hardware. |
| Support for multiple processor architectures (Snapdragon X, Intel Lunar Lake, AMD Ryzen AI 9). | Limited initial availability of models (starting with 1.5B variant). |
| Access to AI Toolkit for developers to experiment with DeepSeek R1 locally. | Concerns over proprietary code claims from competitors like OpenAI. |
| Integration with Azure AI Foundry for cloud-based AI model access. | The rapid development pace may lead to stability or support issues. |
| Utilization of advanced quantization techniques for efficient processing. | Dependence on specific hardware requirements may limit user base. |
| Key positive aspects of Copilot+ | Key concerns related to Copilot+ |
|---|---|
| Double the battery life compared to previous models | Glossy display finish can cause reflections in bright environments |
| Impressive TOPS for AI workloads | Screen brightness may not be sufficient to combat reflections |
| Stylish and ultra-light design, weighing around 1.2 kilos | Shorter key travel on the keyboard may take adjustment for heavy typists |
| Built with Ceraluminium, an environmentally friendly material | AI-driven features still feel early in development and may improve over time |
| Solid chassis that meets MIL-STD 810H military standards | Lack of ASUS’s suite of AI apps like Muse or StoryCube in this model |
| Decent selection of ports, including Thunderbolt 4 support | Limited bass response in speaker audio |
| High-quality 14-inch 3K Asus Lumina OLED display with 100% DCI-P3 color gamut | Touch interface not well optimized in Windows 11 |
| Supports Asus Pen 2.0 for creative tasks | Effectiveness of AI noise-canceling microphone varies with background noise |
| Smooth and precise haptic trackpad with gesture support |
Source: https://powerup-gaming.com/2025/01/30/asus-zenbook-s-14-review/
This is not a Copilot+ review.
Source: https://nextbigwhat.com/microsoft-to-introduce-deepseek-r1-ai-models-to-copilot-pcs/
This is not a Copilot+ review.
Source: https://businesstech.co.za/news/industry-news/809038/asus-announces-all-new-zenbook-a14/
This is not a Copilot+ review.
This is not a Copilot+ review.
Source: https://www.zdnet.com/article/windows-11-24h2-plagued-by-yet-more-bugs-in-2025/
This is not a Copilot+ review.
Source: https://indianexpress.com/article/technology/tech-reviews/dell-xps-13-9350-review-9804039/
This is not a Copilot+ review.
Source: https://www.fonearena.com/blog/445140/microsoft-deepseek-r1-azure-ai-foundry-github.html
| Key positive aspects of Copilot+ | Key concerns related to Copilot+ |
|---|---|
| NPU-optimized versions of DeepSeek R1 enhance AI performance on Copilot+ PCs. | Potential compatibility issues with dynamic input shapes and behavior in the int4 variant. |
| Efficient on-device AI application development due to powerful NPUs. | The complexity of implementing the ONNX QDQ format across various NPUs. |
| Low memory footprint and fast inference achieved through advanced quantization techniques. | Limited support for dynamic tensor operations in the hardware stack. |
| Users can interact with advanced AI models entirely locally, enhancing privacy and speed. | The need for developers to adapt to new AI Toolkit features and models. |
| Availability of multiple model variants (7B and 14B) for diverse application needs. | Potential learning curve for developers unfamiliar with the new AI Toolkit and its functionalities. |