Microsoft Launches Mu AI Model for Windows, Boosts Local Processing

Microsoft Corporation has officially launched its newly developed Mu small language model, an AI tool specifically designed for local operation on personal computers. This move has garnered significant market attention. According to Microsoft's official technical documentation, the model will initially be applied to the "Settings" function of the Windows system, utilizing natural language processing to convert user inputs into system commands. This feature is currently available for testing in the developer preview version of Windows, which includes the Copilot+ function, and is open to participants in the Insider program.

Microsoft has emphasized that the Mu model has achieved deep hardware optimization, with all computational tasks offloaded to the neural processing unit (NPU) for execution. The model's response speed has been tested to exceed 100 tokens per second, meeting the stringent requirements for immediate interactive experiences in system settings. During the model training phase, Microsoft used NVIDIA's A100 data center GPUs to build the training cluster and relied on the Azure machine learning platform to complete the full process of algorithm optimization.

To ensure maximum performance on end devices, Microsoft has collaborated deeply with three major chip manufacturers: AMD, Intel, and Qualcomm. For different NPU architectures, customized quantization computation schemes have been developed. This cross-industry collaborative strategy not only ensures the model's compatibility with various terminal devices but also significantly enhances energy efficiency through hardware acceleration technology.