exo
Build your own AI cluster at home using everyday devices.
- Support popular models such as LLaMA.
- Dynamic model segmentation, based on the optimal allocation model of network topology and device resources.
- Automatic device discovery, no manual configuration required.
- Provide APIs compatible with ChatGPT to simplify the model running process.
- Supports multiple inference engines, such as MLX, tinygrad, llama.cpp.
- Supports multiple network modules, including GRPC, Radio, and Bluetooth.
Product Details
Exo is an experimental software project aimed at utilizing existing devices at home, such as iPhone, iPad, Android, Mac, Linux, etc., to unify into a powerful GPU to run AI models. It supports multiple popular models, such as LLaMA, and has dynamic model segmentation capabilities that can optimally segment models based on the current network topology and device resources. In addition, exo also provides an API compatible with ChatGPT, allowing for the use of exo to run models in applications with only one line of code change.
