I don't understand what you're saying? An LLM is a transformer model trained on a large corpus of natural language, often with some post-training. An image model is a different type of transformer model. What's controversial here?