User runs Hermes on Android with Graphene OS and remote Llama.cpp backend

A user demonstrates running the Hermes interface on Android via a native Debian Terminal within Graphene OS, utilizing a remote backend for inference.

• The setup pairs a mobile RTX 5070 Ti with an eGPU containing an RTX 3090 over Thunderbolt 5 to provide approximately 34GB of VRAM.
• Inference is handled by Llama.cpp and the Qwen 3.6 35b model, achieving 2500 tps for prompt processing and 80-150 tps for generation with MTP.
• The configuration allows for local voice dictation and a sleek GUI on the mobile device while relying on external hardware for compute.