Last updated March 26
Today we’re introducing Gemini 2.5, our most intelligent AI model. Our first 2.5 release is an experimental version of two.5 Pro, which is state-of-the-art on a wide selection of benchmarks and debuts at #1 on LMArena by a big margin.
Gemini 2.5 models are considering models, able to reasoning through their thoughts before responding, leading to enhanced performance and improved accuracy.
In the sector of AI, a system’s capability for “reasoning” refers to greater than just classification and prediction. It refers to its ability to research information, draw logical conclusions, incorporate context and nuance, and make informed decisions.
For a very long time, we’ve explored ways of creating AI smarter and more able to reasoning through techniques like reinforcement learning and chain-of-thought prompting. Constructing on this, we recently introduced our first considering model, Gemini 2.0 Flash Considering.
Now, with Gemini 2.5, we have achieved a brand new level of performance by combining a significantly enhanced base model with improved post-training. Going forward, we’re constructing these considering capabilities directly into all of our models, in order that they can handle more complex problems and support much more capable, context-aware agents.
