exo

Software that joins your computers, such as several Macs, into one AI cluster to run models too large for any single device.

48.1k stars, 3.6k forks Python On GitHub since June 2024

View on GitHub How to run it

What is exo?

exo connects devices on your network into a cluster for running language models. Devices discover each other automatically, and exo splits a model across them based on each machine's memory and the speed of the links between them. With RDMA over Thunderbolt 5 it cuts latency between Macs sharply, and tensor parallelism speeds up inference as you add devices: the authors report up to 3.2 times faster on four devices. It is developed by exo labs.

Good fit if

  • You own several Apple silicon Macs and want to run a frontier-size open model at home.
  • You want to experiment with distributed inference without a GPU server.

Look elsewhere if

  • One machine is enough for your models: Ollama or llama.cpp.
  • You need production serving for many users: vLLM.

Before you adopt it

exo moved from GPL-3.0 to Apache-2.0 in December 2025, a more permissive license. Builds are published for macOS only. In the last 90 days every commit came from a single developer, and the last release was in April 2026.

Drafted with AI from the project's README and repository data, then checked against that data on 30 October 2026. Report a correction

exo health: is it still maintained?

43 of 100, actively developed

  • Activity
    109 commits in the last 90 days, the latest on 29 October 2026.
    25 / 30
  • Contributors
    1 person wrote the latest 100 commits; the most active wrote 100% of them.
    2 / 20
  • Releases
    11 releases in the last 12 months; the latest, v1.0.71, on 23 April 2026.
    11 / 20
  • Replies to issues
    Maintainers answered 0% of recent issues.
    0 / 20
  • Security practices
    No OpenSSF Scorecard for this repository yet, so this part gets a neutral score.
    5 / 10

Read from GitHub on 1 November 2026. How the score works

exo license: is it open source?

exo uses Apache-2.0, an open-source license approved by the OSI.

  1. June 2024First published on GitHub under GPL-3.0.
  2. 17 December 2025Changed to Apache-2.0. Commit: prep repo for v1
  3. TodayApache-2.0.

How to run exo yourself

Docker
No Docker files in the repository
Compose file
None in the repository
Setup effort
Not rated: no compose file to judge from
Kubernetes
No Helm chart in the repository
Desktop and mobile
Builds for macOS in GitHub releases
Interface languages
English only

Install instructions: the README on GitHub.

Alternatives to exo

  • KoboldCppKoboldCpp - Run GGUF AI models locally with a KoboldAI UI. One File. Zero Install. 93 health · AGPL-3.0 · 12.1k stars
  • Page AssistUse your locally running AI models to assist you in your web browsing 74 health · MIT · 8.3k stars
  • AnythingLLMStop renting your intelligence. Own it with AnythingLLM. Everything you need for a powerful local-first… 94 health · MIT · 67.8k stars
  • RAGFlowRAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG… 93 health · Apache-2.0 · 93.3k stars
  • OllamaGet up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models. 89 health · MIT · 186k stars

All 36 projects in AI chat & local LLMs

Questions about exo

Is exo open source?

Yes. exo is published under Apache-2.0, an OSI-approved open-source license.

Is exo still maintained?

Its health score is 43 of 100 (actively developed). It had 109 commits in the last 90 days, the latest on 29 October 2026, and its latest release came out on 23 April 2026.

Did exo change its license?

Yes. It moved from GPL-3.0 to Apache-2.0 on 17 December 2025.

How do I run exo?

You can run it as an app for macOS. The README on GitHub has the exact steps.

What are the alternatives to exo?

Close alternatives in the AI chat & local LLMs category include KoboldCpp, Page Assist, AnythingLLM.