exo
Your spare Macs, united into one AI cluster
What it is
Auto-discovers the machines on your network, shards a model across them with tensor parallelism, and — as of the 2026 relaunch — supports RDMA over Thunderbolt 5 on Apple Silicon for near-datacenter interconnect speeds. Exposes OpenAI-, Claude-, and Ollama-compatible APIs.
Why it's interesting
The most credible open path to running frontier-scale open models at home: two Mac Studios over TB5 instead of an H100 box. The 1.x rewrite turned the pitch from aspirational to real.
Use cases
- Running models too large for any single machine you own
- Pooling office Macs into a private inference cluster
- A local API endpoint backed by distributed inference
Who it's for
Hardware enthusiasts, small teams with several Macs, homelab builders
Setup
Moderate. Best on Apple Silicon (M3/M4 with Thunderbolt 5 for RDMA); uv/Python, Node 18+, Rust nightly. Linux is CPU-only for now
Limitations & cautions
Linux GPU support is still cooking, the RDMA path needs matched macOS versions and TB5 cabling, and clustering adds failure modes single-box tools never see.
Editorial takeaway
Datacenter tricks arriving in living rooms. The gap between 'my hardware' and 'their hardware' keeps shrinking.