Even more so where is the model that is beating GPT-5? This level that fell flat should have been easy to jump over if the scaling narratives were holding.