Google DeepMind Unveils Gemini 2.5 as Its Smartest AI Model Yet
Google DeepMind introduces Gemini 2.5 Pro Experimental, a new thinking model that achieves state-of-the-art results on major AI benchmarks. The model is currently available for developers and Gemini Advanced users.
Google DeepMind introduces Gemini 2.5, marking it as the company's most intelligent AI model to date. The initial release, an experimental version of Gemini 2.5 Pro, claims the top spot on the LMArena leaderboard by a significant margin. This achievement highlights the model's advanced capabilities and high-quality output style based on human preferences.
Unlike traditional AI systems, Gemini 2.5 operates as a "thinking model" that reasons through its internal thoughts before delivering a final response. This approach allows the system to analyze information, draw logical conclusions, and incorporate context more effectively. As a result, the model exhibits strong performance in complex coding, math, and scientific benchmarks.
Developers and users can access Gemini 2.5 Pro Experimental immediately through Google AI Studio and the Gemini app for Advanced subscribers. The company plans to bring the model to Vertex AI soon and release official pricing details in the coming weeks. Moving forward, Google aims to build these integrated thinking capabilities directly into all future Gemini models to support more advanced, context-aware agents.