Proprietary models like Anthropic's Claude Mythos 5 and OpenAI's GPT-5.6 Sol lead in overall benchmark performance, though their exact parameter counts are often undisclosed.
Open the full topic