Microsoft has announced a significant shift in the accessibility of its local Language Model APIs for Windows 11, allowing users to run these APIs on non-Copilot+ PCs, provided they meet new hardware specifications. The primary requirement is an NVIDIA RTX 30 series GPU with at least 6GB of VRAM. This development diminishes the competitive edge previously held by Copilot+ PCs, which debuted on June 18, 2024, and have been instrumental in boosting sales for PC manufacturers.
While the introduction of Copilot+ PCs was initially linked to the presence of a Neural Processing Unit (NPU), it appears that the allure of these machines was more about branding than technical necessity. The NPU, designed for efficient AI model execution, was once seen as essential for running local AI capabilities. However, the raw processing power of GPUs has always positioned them as superior for handling AI tasks, leading to speculation that Microsoft may eventually eliminate the NPU requirement altogether.
To qualify as a Copilot+ PC, a device must feature 16GB of RAM, an SSD, and a minimum of 40 TOPS NPU. Despite these specifications, the market has seen a surge in sales of what are now branded as “Copilot+ PCs,” even for standard laptops, reflecting a growing consumer interest in AI capabilities.
In a recent update on GitHub, Microsoft confirmed that local AI functionalities, previously exclusive to Copilot+ PCs, are now accessible to a wider range of devices equipped with compatible GPUs. Developers can utilize the Language Model APIs on non-Copilot+ PCs, effectively broadening the scope of local AI capabilities across Windows 11 devices.
What are the potential new AI features coming to all PCs with GPUs via Language Model APIs?
Exploring Microsoft’s official documentation reveals how this API unlocks new AI features for all PCs equipped with capable GPUs. The Language Model APIs facilitate direct access to AI functionalities within Windows applications, powered by a compact language model known as “Phi Silica.” This model has already been referenced in Microsoft Edge for features such as “Rewrite using AI.”
Currently, Windows PCs do not come preloaded with AI models unless they are Copilot+ PCs. However, with the new Local Model APIs, applications can prompt Windows Update to download the Phi Silica model, enabling local execution via the GPU. The following AI features are now accessible through this integration:
- AI-powered text formatting in applications using Windows.AI.Text APIs
- Summarization capabilities (TextSummarizer)
- Text rewriting (TextRewriter)
- Text-to-table conversion (TextToTableConverter)
- General prompt generation
This means that users with NVIDIA GPUs will soon be able to utilize ChatGPT-like functionalities directly within Windows applications, all while ensuring their privacy through local processing, as opposed to relying on cloud-based models like Copilot or ChatGPT. However, it is important to note that certain features, such as Windows Recall and Click to Do, remain exclusive to devices with NPU chips, indicating that while the landscape is evolving, some limitations still persist.
