Gemini is a set of large language models (LLMs) developed by Google AI that leverages training techniques from AlphaGo, including reinforcement learning and tree search.