Google has recely unveiled a new artificial ielligence model called Jamna 2.5 Pro, which is claimed to be the most advanced achieveme in the Jamna collection. The model is designed to do sophisticated tasks and uses powerful reasoning abilities that Google calls Thinking Built-in. Those who use the Gemini Advanced service can now work with this model.
According to Google, the Jammeh 1.5 series is naturally reasonable. This feature allows the model to examine the data, achieve logical results, and to better understand the difficult challenges with a better understanding of the issues. This ability, called “inner thinking”, is prese in all the models of this series, though they are no longer used in their naming. Users can see the steps of this process by activating the Show Thinking option in the Jammeh program.
The Gemini 2.5 Pro model, known as the NEBULA password and the official Gemini-2.5-EXP-03-25, is the first member of the 5.1 Series and makes significa progress compared to the previous models. Google has announced that the model has reached the ranking of LMARNA, which is based on human opinion on the quality of responses.
Peru” width=”1448″ height=”2000″ loading=”lazy”/>
The model has also been successful in two specialized tests. In Aime 2025, which includes hard mathematical questions at the Olympiad level, and in GPQA Diamond, which measures the ability to scieific reasoning and answer sophisticated questions, Jammeh 2.5 Peru has the highest score. It is noteworthy that this success is achieved without the use of methods such as “Majority Voting” or Majority Voting, a method in which several answers are produced and the most promine.
In the Humanity’s Last Exam series, the model also reached 4.9 % without the help of foreign tools. Google considers this result to be the best recorded among models that are measured only with their iernal capabilities. This benchmark, which assesses knowledge and reasoning in more than a hundred differe topics, offers accurate and diverse questions from the humanities, natural and analytical sciences.
Gemini 2.5 Pro advances are the result of fundameal changes in the main structure of the model and improvemes in post -developme education. This model is made to solve more difficult problems and support ages that require a deeper understanding of text and coext. In the field of programming, this model has also made a big difference. Google says the Gemini 2.5 Pro has better coding capability than version 4.3 and offers reliable performance in tasks such as building web applications, designing coding tools and editing code. In the Swe-Bench Verified benchmark, which measures automatic code production, this model has achieved a score of 4.9 % with the help of a special age.
Google’s new model supports a one -million -token background window and will soon increase to two million token. This feature allows the model to process a huge amou of information, such as the eire coding tanks, together. In addition, the Gemini 2.5 Pro can analyze data in various formats such as text, sound, image and video.
Currely, access to this model is possible through Gemini Advanced and Google AI Studio and will be added to the Vertex AI in the coming weeks, Google’s cloud cloud platform for the developme of organizational artificial ielligence models. Google will soon release more information on pricing and capacity enhancing options for large projects.
Source: 9to5Google
