21 results found
A well-known prompt used to test AI geography skills no longer works on the O3 model, prompting debate about benchmark reliability and model drift.
