Comment by nialv7

There are so many AI generated inference engine for local models now, each of them are generally narrower but they are all faster than llama.cpp. Maybe llama.cpp needs to rethink their strategies...