June 2, 2025
Microsoft introduces Phi-3, a family of lightweight AI models designed to run efficiently on phones, laptops, and edge devices with impressive performance.
Microsoft has launched Phi-3, a family of small language models (SLMs) aimed at bringing advanced AI capabilities to devices with limited resources. The models are designed to be highly efficient and accurate, making them ideal for deployment on smartphones, tablets, and edge computing platforms.
This move is part of Microsoft’s broader strategy to support AI at the edge, allowing developers to leverage generative AI features even without constant access to the cloud.
The Phi-3 models come in three sizes:
Phi-3-mini (3.8B parameters)
Phi-3-small (7B parameters)
Phi-3-medium (14B parameters)
These models are trained using a unique dataset focused on high-quality, textbook-style content, synthetic data, and curated instructional materials. According to Microsoft, this enables Phi-3 to outperform models double its size on key benchmarks.
