🤖 GPT-5.2 has nearly reached human-level reasoning — the AI scored 53–54% on the most difficult AI test, ARC-AGI-2, while the average human score is 60%.
Additionally, GPT-5.2 perfectly solved one of the hardest math exams, AIME 2025, for the first time and achieved a record 70–74% on the GDPval test, which measures “real-world human work,” often reflecting the level of a strong specialist.
#Aİ
