OpenAI Claims Its New Model Reached Human Level on a Test for ‘General Intelligence.’ What Does That Mean?

OpenAI’s o3 system scored 85% on the ARC-AGI benchmark, well above the previous AI best score of 55% and on par with the average human score.