Where is Multi-Head Attention used in practice?

Most large modern LLMs use GQA rather than full multi-head attention. AI governance teams document the attention configuration (number of heads, head dimension, GQA grouping) as part of model architecture lineage. T