Gemini 2.5: Our Smartest AI Model Yet

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 2.5 Pro, which is state-of-the-art on a wide range of benchmarks and debuts at #1 on LMArena by a significant margin.

Gemini 2.5 models are designed for enhanced reasoning capabilities, allowing them to analyze their thoughts before responding. This leads to superior performance and improved accuracy.

In the realm of AI, a system’s capacity for “reasoning” extends beyond mere classification and prediction. It encompasses the ability to analyze information, draw logical conclusions, incorporate context and nuance, and make informed decisions.

For a long time, we have investigated methods to enhance AI’s reasoning abilities through techniques like reinforcement learning and chain-of-thought prompting. Building on these advancements, we recently introduced our first thinking model, Gemini 2.0 Flash Thinking.

With Gemini 2.5, we have reached a new level of performance by integrating a significantly enhanced base model with improved post-training techniques. Going forward, we are committed to embedding these advanced reasoning capabilities into all of our models, equipping them to tackle more complex problems and support even more capable, context-aware agents.