Note: The generated text is optimized for token count. GPT-4o and Claude tokenize text slightly differently, so approximate token counts are used.