AI AgentsGrok 4.5 Is Cheap Enough That the Benchmark Gap Might Not MatterGrok 4.5 changes the model-routing question: when does a cheaper near-frontier model beat the benchmark leader on cost per accepted task?5 days ago