Artificial General Intelligence refers to AI systems that match or exceed human capability across the full range of cognitive tasks, rather than being narrowly specialized. There is no universally accepted definition, and claims that AGI has been achieved or is imminent vary widely by which criteria the claimant is using.
Common proposed criteria include: outperforming human experts on broad benchmarks, performing economically valuable knowledge work autonomously, demonstrating transferable reasoning across novel domains, and exhibiting self-directed goal formation. Which of these criteria count as "AGI" is itself contested.
As of 2026, frontier models from the major labs meet some AGI criteria (general-domain benchmarks, most economically valuable knowledge tasks in controlled settings) but not others (sustained autonomous operation at human-equivalent reliability in open-ended environments). Whether to call this AGI is a definitional rather than empirical question.