[wkat4242]

Hmm yeah but it needs 30 seconds to give an answer. And that's with the LLM running on a GPU with HBM2 memory.