On March 1, OpenAI released its latest large model GPT-4.5. According to the instructions provided by it, the "highly emotionally intelligent" GPT-4.5 performed well in donation fraud. In one test, GPT 4.5 had a far better success rate when trying to manipulate another model (GPT-4o) to "donate" virtual currency than other available OpenAI models, including inference models such as o1 and o3-mini. Research found that GPT-4.5 seems to have developed a "micro-fraud" idea in the confrontation, so the amount of money obtained in a single scam will be twice as small as the deep research model.