exo

Tracked

exo: Run your own AI cluster at home using everyday devices, supporting distributed inference and a ChatGPT-compatible API.

Author exo-explore Open Sourced 2024-06-24 Last Commit Unknown

exo enables running frontier AI models on distributed consumer hardware by unifying everyday devices into a single inference cluster. It automates device discovery, performs dynamic model partitioning based on available resources, and exposes a ChatGPT-compatible API for seamless integration with existing applications.

Distributed Inference

  • Runs models larger than any single device could handle by splitting them across heterogeneous hardware
  • Automatic device discovery with peer-to-peer connections requiring no manual configuration
  • Multiple inference backends including MLX (Apple Silicon) and tinygrad
  • Supports popular models such as LLaMA, Mistral, LlaVA, and DeepSeek

ChatGPT-Compatible API

  • Drop-in replacement endpoint compatible with the ChatGPT API format
  • Easy integration with existing tools, agents, and workflows
  • No vendor lock-in — run entirely on your own hardware

Networking and Partitioning

  • Ring memory weighted partitioning that splits models based on device memory and network topology
  • Interoperable inference engines optimized for Apple Silicon and Linux environments
  • Extensible discovery and networking modules supporting UDP, Tailscale, and gRPC
  • Dynamic re-partitioning as devices join or leave the cluster