LLMs are fundamentally the same since the first GPT dropped. They are suffering from the same defects and nothing apart fro mdataset size has changed, as these defects(hallucinations etc) are inherent to the design and cannot be solved. LLMs are a dead end.