QVAC Fabric enables distributed AI model execution across devices

Gianfranco: Gianfranco is an individual associated with the QVAC project who conducted testing and demonstration of its distributed capabilities. In the news, Gianfranco executed a large 144 GB model across connected DGX Spark machines using QVAC Fabric's new cluster inference feature. QVAC Fabric: QVAC Fabric is the core inference and fine-tuning engine in the QVAC ecosystem by Tether, forked from llama.cpp to support local AI workloads on consumer hardware including mobile devices and desktops via cross-platform backends like Vulkan and Metal. It enables on-device model execution, LoRA fine-tuning, and related capabilities while keeping data private and offline. In this news, QVAC Fabric introduces support for splitting a single model across multiple machines, demonstrated with cluster inference for larger models like DeepSeek V4 Flash. Use Case: Cluster inference allows users with multiple devices to run significantly larger AI models locally at home. Product Update: QVAC Fabric now supports distributed cluster inference to split models across multiple connected machines.