The Anthropic team's release of the Claude 3 series models has surpassed OpenAI's GPT-4. Comprised of former OpenAI members, they focus on researching large-scale model technology, committed to developing safe, useful, and harmless large models. They enhance the capabilities of large models through RLHF technology and introduce the concept of "constitutional artificial intelligence" for value cultivation.