All open source
Open SourceLocal46k

exo

Connect everyday devices into a p2p cluster to run frontier models.

exo is an open-source project for running large AI models distributed across multiple everyday devices, such as laptops, phones, and single-board computers. It dynamically partitions a model across a peer-to-peer cluster of heterogeneous hardware, letting devices combine their memory and compute to run models too large for any single machine. It exposes an OpenAI-compatible API for the unified cluster.

Repository

exo-explore/exo

Language

Python

What you'd build with it

  • Running a large model across several consumer machines instead of buying one high-VRAM GPU
  • Pooling memory from mixed devices on a local network to fit bigger models for inference
  • Experimenting with distributed, peer-to-peer LLM serving using existing hardware

Tags

distributedclusterlocal