The model also improved performance on competitive programming questions and exceeded human PhD-level accuracy on a benchmark of science problems, the company said.
OpenAI CEO Sam Altman touted the new model series in a post on X, calling them "our most capable and aligned models yet."