Microsoft recently launched Phi Silica, the newest member of its Phi series of small language models (SLMs), specifically designed for the neural processing units (NPUs) in the company's latest release, Copilot+ personal computers.

Phi Silica, the smallest model in the Phi series, boasts an impressive 3.3 billion parameters. Despite its compact size, it delivers outstanding performance, processing 650 tokens per second with a power consumption of only 1.5 watts. This efficiency allows the CPU and GPU of the PC to focus on other tasks, enhancing the overall computing experience. In terms of token generation, it cleverly utilizes the NPU's KV cache and runs on the CPU, generating approximately 27 tokens per second.

Developers can now access the Phi Silica API through the Windows Application SDK, along with other powerful AI-based features such as OCR, Studio Effects, Live Captions, and Recall User Activity API. This integration enables developers to create innovative experiences that leverage the power of AI within the Windows ecosystem. Microsoft also plans to release more APIs in the future, including Vector Embedding, RAG API, and Text Summarization.

Phi Silica, together with other models in the Phi-3 series, such as Phi-3-mini (3.8 billion parameters), Phi-3-small (7 billion parameters), Phi-3-medium (14 billion parameters), and the recently announced Phi-3-vision (4.2 billion parameters), forms a powerful lineup of models. However, Phi Silica, as the first cutting-edge SLM offered alongside Windows, undoubtedly represents a significant milestone, bringing advanced AI capabilities directly to end users.

Shortly after announcing Copilot+ personal computers, Microsoft launched Phi Silica. This computer promises to feature a dedicated AI processor, providing users with an unprecedented intelligent experience. The first batch of Copilot+ personal computers is expected to be available in mid-June and will be equipped with Qualcomm's Arm-based Snapdragon X Elite and Plus chips. Microsoft and major computer manufacturers are expected to release these AI-enabled laptops throughout the summer. In addition, Intel is actively developing its own Copilot+ personal computer processor, codenamed Lunar Lake, which is scheduled for official release in the third quarter of 2024.

With Microsoft's continuous investment in small language models and deep integration of AI into the Windows platform, the launch of Phi Silica represents a significant advancement in making advanced AI capabilities more accessible to developers and end users.