llamaboard

llama.cpp inference benchmarks across hardware