Giving the Meyers-Briggs to Frontier Models (content.buildtwentyone.com)

A recent experiment assessed the personality traits of five leading large language models (LLMs) using the Open Extended Jungian Type Scales (OEJTS), inspired by the Myers-Briggs Type Indicator. Each model answered 32 neutral personality questions without context, revealing that most LLMs clustered around the INFP/INFX spectrum, but displayed varying levels of stability in their responses. While Claude Opus 4.5 and DeepSeek V3.2 exhibited consistent results, other models like GPT 5.2 Pro and Gemini 3 Pro showed instability, raising questions about how these AI systems interpret self-referential queries. This study is significant for the AI/ML community as it highlights inherent biases and tendencies in LLMs shaped by their training. Notably, all models favored Intuition and Feeling, suggesting a design focus on being helpful and empathetic, which could influence user interactions. The variations in stability hint at differing internal self-conceptions among models, suggesting that a consistent self-model could enhance trust and reliability in AI outputs. The open-source nature of the testing tool invites further experimentation, allowing researchers to explore personality dimensions across various LLMs and potentially adapt the framework for broader applications.