GPT-5 Can Only Complete 17% of Spatial Tasks That Humans Find Routine — A New Benchmark Quantifies the Gap

GPT-5 Can Only Complete 17% of Spatial Tasks That Humans Find Routine — A New Benchmark Quantifies the Gap | BedrockNews