We ran 13,825 personality evaluations on 6 LLMs (GPT-5.2, Claude Opus 4.5, Llama 70B/8B, Mistral Large 3, Qwen 72B) and found that open-weight models cluster together with nearly identical personality profiles, while closed frontier models have diverged into distinct types.

Surprisingly, Llama 8B and 70B score within 0.7 points of each other across all 10 dimensions, suggesting personality is shaped more by training methodology than model scale.