Sadly, it seems Meta made some good decisions in LLM area at least. Llama is a de facto base of most good 7b and 13b LLMs, the ones that have potential on running on low grade hardware.