✏️ Results of Google Gemini 3.0 vs. OpenAI ChatGPT (5.2) CSAT Competition
OpenAI countered with the release of Google's "Gemini 3.0" model with its "ChatGPT 5.2" model.
1️⃣ ChatGPT 5.2: "Math Master"
Perfect Score in Math: It demonstrated overwhelming performance in math, solving Geometry Problem 29, considered a difficult problem, in just four minutes.
Overall Score: It outperformed Gemini 3.0 in four out of six subjects.
The "Deep Thinking" feature is believed to have increased computation time and improved accuracy.
2️⃣ Gemini 3.0: "Strong in Visual Information and Reasoning"
Perfect Score in Korean History: It demonstrated outstanding "multimodality" skills, integrating images and diagrams with text to make inferences. Its visual interpretation skills were excellent, as evidenced by its ability to accurately predict subsequent events from photographs of captured US troops during the Shinmiyangyo campaign.
Weaknesses Improved: Image interpretation capabilities have significantly improved, even to the point of understanding illustrations in ancient documents.
3️⃣ Model Limitations
ChatGPT: Even when solving problems successfully, it sometimes failed to understand complex text containing mixed text and tables, leading to incorrect answers.
Status of Domestic AI: Due to its performance lagging behind that of international models, some argue that it should be differentiated from general-purpose AI by focusing on specialized models.
