GPT-5.6 can already do as well as an intern, but the evaluation criteria are still vague—using the amount of GPU resources put in to assess it is actually a pretty solid approach.

CoinWorld news, OpenAI senior researcher Noam Brown said that if it were in most AI research tasks today, he would choose GPT-5.6 over a human research intern. Brown made this statement while responding to "Has AI reached the level of an AI research intern?" during ICML 2026. He also acknowledged that there is currently no accepted method to measure whether AI has truly reached the level of a human research intern. Another OpenAI researcher, Yash Pande, proposed a more practical measurement: how many GPUs researchers are willing to allocate to Codex to complete research tasks. If AI is truly stronger than humans, the company would theoretically invest more computing power.