typically, when any AI company says a model performs as well as fable, all they're really telling us is that the benchmarks that exist for measuring AI capabilities aren't very good.