Six Large Language Models outperformed humans on five ability emotional intelligence tests. ChatGPT-4 also successfully generated new test items for each test, with the AI-created versions showing psychometric properties similar to the originals.