Sat Mar 29 12:30:00 UTC 2025: ## Google DeepMind Unveils Gemini 2.5: Its “Most Intelligent AI Model” Yet
**London, UK** – Google DeepMind has announced Gemini 2.5, its latest and most advanced AI model to date. The company claims Gemini 2.5 Pro, an experimental version of the model, achieves state-of-the-art results across numerous benchmarks, showcasing significantly improved reasoning capabilities.
According to Google DeepMind CTO Koray Kavukcuoglu, Gemini 2.5 models are “thinking models,” capable of analyzing information, drawing logical conclusions, and incorporating context before generating responses. This “reasoning” surpasses simple classification and prediction, enabling more informed decision-making. These advancements build upon previous work using reinforcement learning and chain-of-thought prompting.
Gemini 2.5 Pro has already achieved top rankings on several key leaderboards, including LMArena, demonstrating superior performance and high-quality output. It particularly excels in math and science benchmarks (GPQA and AIME 2025) and achieved a state-of-the-art score of 18.8% on Humanity’s Last Exam, a challenging test of human-level knowledge and reasoning. Impressive coding abilities are also highlighted, with Gemini 2.5 Pro achieving a 63.8% score on SWE-Bench Verified and demonstrating the capability to generate executable video game code from a single-line prompt.
The model boasts a one-million-token context window (expanding to two million soon), allowing it to process vast datasets and diverse information sources, including text, audio, images, video, and code. Gemini 2.5 Pro is currently available for experimentation in Google AI Studio and to Gemini Advanced users, with a planned rollout to Vertex AI in the coming weeks. Google DeepMind encourages user feedback to further refine the model’s capabilities. The release marks a significant leap forward in AI reasoning and problem-solving, paving the way for more sophisticated and context-aware AI applications across various industries.
First Piper 