• by Laurie Sullivan  @lauriesullivan, March 27, 2025

Google has released what it calls a thinking model that takes users into its thought process to reveal assumptions and why it chose the line of reasoning for its responses.

"In the field of AI, a system’s capacity for 'reasoning' refers to more than just classification and prediction," Koray Kavukcuoglu, chief technology officer at Google DeepMind, wrote in a blog post. “It refers to its ability to analyze information, draw logical conclusions, incorporate context and nuance, and make informed decisions.

Gemini 2.5 Pro Experimental is a multimodal reasoning AI model that became available on Tuesday in the company’s developer platform, Google AI Studio, as well as in the Gemini app for subscribers to the company’s monthly AI plan, Gemini Advanced. It soon will become available in Vertex AI.

In the future, Kavukcuoglu wrote, Google will build these thinking capabilities directly into all of its models, so they can handle more complex problems and support even more capable, context-aware agents. Thinking models are capable of reasoning through their thoughts before responding, resulting in enhanced performance and improved accuracy.

Since this model is experimental, it may not always work as expected, and some features will not be available, Kavukcuoglu explained.

Gemini 2.5 Pro also announced this week tops the LMArena leaderboard — a crowdsourced, real-time ranking system for Large Language Models (LLMs). It measures human preferences, indicating a highly capable model equipped with high-quality style.

The model, 2.5 Pro, also shows strong reasoning and code capabilities, leading on common coding, math and science benchmarks.

As an example in the post, Google provides an example of how 2.5 Pro can use reasoning capabilities to create a video game by producing the executable code from a one-line prompt.

Google did not publish API pricing for Gemini 2.5 Pro, but said it expects to providing pricing in the coming weeks.