WWDC26: Explore distributed inference and training with MLX | Apple
Apple's MLX now supports distributed LLM inference and training across multiple Macs via RDMA over Thunderbolt 5 and the new JACCL library.
“JACCL is an open-source collective communication library built by Apple. It leverages RDMA over Thunderbolt and gives you collective communication primitives for sending data between machines and combining results across the group”
At WWDC26, Apple's MLX team detailed how to cluster multiple Apple Silicon Macs for distributed LLM inference and fine-tuning, enabled by new RDMA-over-Thunderbolt-5 support in macOS 26.2 and Apple's open-source JACCL communication library. This pushes local, on-device AI past single-machine memory and compute limits, signaling Apple's continued investment in private, hardware-native AI workloads as an alternative to cloud inference.