Last updated March 26
Today we introduce Gemini 2.5, our most intelligent AI model yet. Our first version 2.5 is an experimental version of version 2.5 Pro, which is state of the art in many benchmarks and debuts at number one on the list LMA by a considerable margin.
Gemini 2.5 models are thinking models, able to think through their thoughts before responding, resulting in increased efficiency and greater accuracy.
In the field of artificial intelligence, a system's ability to “infer” refers to more than just classification and prediction. It refers to the ability to analyze information, draw logical conclusions, consider context and nuance, and make informed decisions.
We have long been exploring ways to make artificial intelligence smarter and more capable of reasoning through techniques such as reinforcement learning AND prompting in the form of a thought chain. Building on this, we recently launched our first thinking model, Gemini 2.0 Flash Thinking.
Now with Gemini 2.5 we have reached a new level of performance, combining a significantly improved base model with improved post-training training. In the future, we'll be building these thinking capabilities directly into all of our models so they can tackle more complex problems and support even more efficient, context-aware agents.
