ChatGPT o3 Could Secure AIR 4 in JEE Advanced — Leaves Experts Stunned

An intriguing experiment by an engineer from IIT Kharagpur has revealed that OpenAI’s latest model, ChatGPT o3, could have secured an All India Rank (AIR) of 4 in the JEE Advanced—one of the toughest entrance exams in the world and the gateway to India’s premier engineering institutes, the IITs (Indian Institute of Technology).

The experiment was conducted by Anushka Aashvi, an IIT Kharagpur engineer, who began the project as a casual inquiry. What started as curiosity soon turned into a groundbreaking result: ChatGPT o3 scored 327 out of 360, a score that would place it among the top five rank holders in the actual examination.

Aashvi meticulously simulated real test conditions. In her blog, “Heltar,” she explained that the AI was instructed to behave like a genuine JEE aspirant—solving each question independently without using the internet or external Python tools. To further ensure fairness and eliminate memory bias, each question was posed in a fresh chat session, and no hints or corrections were given throughout the test.

Despite these rigorous constraints, ChatGPT o3 exhibited extraordinary proficiency, achieving perfect scores (60/60) in both Chemistry and Mathematics during the second phase of the simulated exam. It lost only a few marks in Physics and earlier sections.

The experiment has sparked widespread interest in the capabilities of advanced AI systems and raised important questions about the future of education, competitive testing, and how society defines intelligence in the age of artificial intelligence.