Sam Altman says OpenAI will have AGI by the end of 2026 if you accept his definition

Details

OpenAI CEO Sam Altman told TIME magazine that he expects his company to have a system meeting OpenAI's internal AGI definition by end of 2026 — though he acknowledges they are "not quite yet" there. Chief Research Officer Mark Chen estimates OpenAI is "80% of the way" to that milestone, and co-founder Greg Brockman believes historians will mark this period as the moment AGI emerged. The claims come from an extensive TIME investigation by Alex Heath, who spent more than two weeks interviewing over 20 OpenAI executives, employees, investors, and rivals. The centerpiece of this confidence is Astra, OpenAI's upcoming model family. Chief Scientist Jakub Pachocki told TIME that Astra already meets their internal AGI benchmark: given an experimental idea, it can write code in OpenAI's own codebase, run the experiment, and report back — compressing roughly a week of human researcher work. The system also enables "persistent agents," virtual coworkers capable of running autonomous tasks over extended periods. Altman's boldest claim is that Astra will be "the first model where the model actually invents new things in a way that matters," pointing toward recursive self-improvement. External researchers are split — some see early signals, others argue that LLMs still cannot generalize or make genuinely novel discoveries, which many consider baseline requirements for any meaningful AGI definition. Worth noting: OpenAI's AGI definition is explicitly economics-driven ("highly autonomous systems that outperform humans at most economically valuable work"), narrower than many technical framings. Separately, Altman confirmed a "small handful" of hardware devices — table, pocket, and wearable form factors — expected in early 2027.