One key element missing from all these model cards are the model size/number of parameters. Without that info we are in the dark. We can't predict the future of AI. How does the intelligence scale with the increasing #parameters? Is there a limit? Should we attribute incrementally better metrics to larger model size or other techniques? Do they announce the full model they trained or a smaller version that is economically viable for the market conditions? If they double the model size will it be a Professor-level intelligence, a super-human level intelligence or a couple of phds level intelligence?