EXO is a new peer-to-peer cluster runtime that allows users to connect multiple devices into a single AI cluster for running large models locally. It automatically discovers and connects devices on the same network, provided they run the same version of EXO. The system utilizes Apple's MLX framework for inference and supports various API surfaces, including OpenAI Chat Completions and Ollama. EXO also features topology-aware auto-parallelism for efficient model splitting across devices and boasts RDMA over Thunderbolt 5 support for reduced latency between machines. AI
IMPACT Enables local execution of larger AI models by pooling resources across multiple devices.
RANK_REASON The item describes a software tool for creating local AI clusters, not a frontier model release or significant industry event.
- Apple Inc.
- Claude Messages
- EXO
- EXO Labs
- Hugging Face
- MLX
- Ollama
- OpenAI Chat Completions
- OpenAI Responses
- OpenWebUI
- RDMA over Thunderbolt 5
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →