I suspect this is more marketing than fact. I don't think there's any real evidence that this is the case, just hints dropped by OpenAI and Anthropic.
Well, except the giant dump of math problems they just solved with an internal model.
Well, except the giant dump of math problems they just solved with an internal model.