Microsoft is enhancing its Phi line of open-source language models by introducing Phi-4-mini and Phi-4-multimodal. Phi-4-mini, with 3.8 billion parameters, is designed for mobile devices and employs a decoder-only transformer architecture that optimizes hardware usage. It demonstrates superior performance in tasks requiring complex reasoning. Phi-4-multimodal, an upgrade, features 5.6 billion parameters and can process text, images, audio, and video. Both models are reported to outperform competitors in specific tasks, emphasizing their efficiency and capabilities in multimodal processing.
Phi-4-mini contains 3.8 billion parameters, making it compact enough to run on mobile devices. It's based on a decoder-only transformer, analyzing only preceding text.
The Phi-4-multimodal model, with 5.6 billion parameters, advances upon the mini model by processing text, images, audio, and video, enhancing its capabilities.
Collection
[
|
...
]