[gigel82]

Like InternVL, no llama.cpp support severely limits its applications. Close to GPT4v performance level and runnable locally on any machine (no need for a GPU) would be huge for the accessibility community.