Sam Altman Says OpenAI Could Reach AGI by the End of 2026

Sam Altman has made one of his most ambitious predictions yet, saying OpenAI expects to have an internal system that qualifies as artificial general intelligence (AGI) by the end of 2026. In a recent TIME interview, Altman acknowledged that OpenAI has not reached AGI yet, but said the company believes it is approaching the milestone rapidly.

OpenAI’s confidence is reportedly tied to its next-generation Astra model family. The company says Astra is being developed to operate more like an autonomous research assistant, capable of writing and running experiments, analyzing research and completing complex technical work with limited human intervention. Chief Research Officer Mark Chen reportedly estimates that OpenAI is about 80% of the way toward its internal AGI target.

The claim is significant because AGI does not have one universally accepted definition. OpenAI has generally described AGI around highly autonomous systems capable of outperforming humans at most economically valuable work. Other researchers use broader or narrower definitions, meaning that whether Astra—or any future model—actually qualifies as AGI would remain a subject of debate.

OpenAI’s progress is also accompanied by major safety concerns. The company has recently acknowledged that its frontier models are becoming capable enough to require stronger safeguards, including additional monitoring and security work. Reports indicate OpenAI temporarily paused some frontier reinforcement-learning training while it strengthened its safety and alignment systems.

If OpenAI achieves its stated goal, the consequences could be enormous for software development, scientific research and knowledge work. But Altman’s prediction is still a company forecast, not an independently verified declaration that AGI has arrived. The real test will be whether OpenAI can demonstrate a system that consistently performs a broad range of complex intellectual tasks at the level its definition requires.