It’s safe to assume that after less than a decade of LLM development, we’re nowhere close to the limit yet. In fact, progress still seems to be accelerating at the moment.

If anything, it should be safe to assume by now that capabilities don't scale linearly with model size.