On 25th March, Google unveiled its most advanced artificial intelligence model to date, called Gemini 2.5. According to the company, this system represents a new generation of models capable of “thinking,” that is, reasoning internally before generating a response, resulting in higher levels of accuracy and optimised performance.
The model, identified as Gemini 2.5 Pro Experimental, has been specifically designed to tackle complex tasks. It tops the LMSys Arena ranking, a platform that evaluates human preferences, with a significant advantage over other models, positioning it as a highly competent AI with a high-quality response style, according to Google.
Think you know Gemini? Think again.
Meet Gemini 2.5: our most intelligent model 💡 The first release is Pro Experimental, which is state-of-the-art across many benchmarks – meaning it can handle complex problems and give more accurate responses.
Try it now →… pic.twitter.com/bFcx0IlY24
— Google DeepMind (@GoogleDeepMind) March 25, 2025In addition, it promises strong reasoning and programming skills, surpassing the usual standards in areas such as mathematics, science, and software engineering.
Gemini 2.5 Pro is now available in Google AI Studio (click to try) and the Gemini app for Gemini Advanced users, with integration into Vertex AI planned for the near future.
What Google’s new AI can do
Gemini 2.5 Pro Experimental, Google’s ‘best’ artificial intelligence model, represents a significant advance in the ability to understand, reason, and generate complex responses.
Designed to solve sophisticated tasks, this model is reported to stand out for its ability to analyse information, synthesise data, and offer precise solutions in multiple disciplines, such as programming, mathematics, and applied sciences.
One of the key new features of Gemini 2.5 Pro is its ability to reason before responding. This allows it to maintain coherence in long conversations, follow detailed instructions, and perform complex comparative analysis.
For example, a user might ask you to analyse a piece of Python code, identify errors, propose an optimised solution, and explain why you suggest that approach, all within a single interaction.
You can also summarise scientific articles, translate technical texts with a high level of accuracy, or create complete presentations from a research paper. In education, you can solve math problems step-by-step, explain advanced physics concepts, or guide students through writing structured essays.
“We’ve focused on coding performance, and with Gemini 2.5 we’ve made a huge leap forward from version 2.0, with more improvements to come. 2.5 Pro excels at creating visually compelling web apps and agentic code applications, as well as code transformation and editing,” Google adds.
What is Gemini Advanced?
Gemini Advanced is a Google subscription plan that offers access to its most advanced artificial intelligence models, including experimental versions not available in other plans.
This subscription is aimed at users who require superior processing capabilities for complex tasks such as in-depth research, advanced programming, and large document analysis.
One of the key benefits of Gemini Advanced is access to the Deep Research feature, which saves time on research projects through smarter searches and detailed information analysis.
This plan also allows you to process and understand files of up to 1,500 pages, including entire books and extensive technical reports.
For developers, Gemini Advanced facilitates more efficient programming by integrating the option to upload code directly to a repository, optimizing collaborative work and automating tasks. Additionally, the subscription includes 2 TB of storage in Google One, allowing large volumes of data to be managed without the need for external services.
The plan also improves integration with the Google ecosystem, offering access to Gemini in applications such as Gmail, Docs, and other productivity tools. This feature is available in select languages, expanding its use in various professional and educational environments.
You can try this new AI model, Gemini 2.5 Pro Experimental, in Google’s AI Studio. Click here to try.