llama.cpp is good for cpu/gpu inferencing
vllm is good for tensor parallelism
ollama sits on top of llama.cpp and shares most of its advantages and shortcomings

#ai #llama #vllm