Google launched a GMINI 2.5 PRO

Google has introduced a new Gmini 2.5 Shi model that “think”. They pause before answering for reflection.

IntRODUKING GEMINI 2.5, OUR MOST INTELLIGENT Ai Model.

Our First Release, An Experimental Version of 2.5 Pro, Unlocks State-of-The-Art Performance in Math and Science. 🔥

Learn more 🧵 pic.twitter.com/aoe7eglijb

– Google (@google) March 25, 2025

The first of the series was issued by the experimental version of Gemini 2.5 Pro. Multimodal neural network with reasoning “is the most intelligent for today.” It is available on a platform for Google AI Studio developers and GEMINI applications for Advanced subscription owners worth $ 20 per month.

“In Gemini 2.5 we have reached a new level of productivity by combining a significantly improved basic model with an improved posting», – said in the blog of the corporation.

The company added that in the future, all its models will have the ability to think.

In the test of AIDER Polyglot, which tracks programming skills, Gemini 2.5 Pro scored 68.6% of points, ahead of leading solutions from Openai, Anthropic and Deeceek.

In the SWE-Bench Verified Benchmark, which measures the ability of software developers, Google’s decision lost Claude 3.7 Sonnet-63.8% against 70.3%.

In the Humanity’s Last Exam multimodal test, consisting of a thousand questions in mathematics, humanities and natural sciences, Gemini 2.5 Pro scored 18.8%, exceeding the results of competitors.

GEMINI 2.5 PRO results compared to other AI models in different benchmarks. Data: Google.

GEMINI 2.5 PRO has a context window of 1 million tokens – the model is able to perceive about 750,000 words at a time. In the future, it is planned to double the indicator.

AI, which “thinks” requires more time and computing resources to give an accurate and spreading answer. The first neural network that “thinks” showed Openai – in September 2024 it released O1. Other SI-developers began to create their own solutions. Among them Deeseek, Anthropic, Alibaba and others.

The gun