- Google Ad Gemini 2.5
- Gemini 2.5 Pro Experimental is available for paid subscribers at the moment
- At the top of the last examination of humanity, the most difficult IA reference
Google has just announced Gemini 2.5, and it is a major upgrade to Gemini that the company calls its “smartest AI model” to date.
Announced on the company’s blog, Google has revealed the experimental version of 2.5 Pro, which is available today for all Gemini Advanced subscribers. More models 2.5 will arrive in the future.
Google Gemini 2.5 models are a new generation of reflection models that are able to achieve “a new level of performance by combining a considerably improved basic model with improved post-training”.
The thought capacities of 2.5 will be implemented in all future Google AI models, which, according to the company, will allow them to “manage more complex problems and to take care of even more competent and aware agents”.
So what does that mean? Well, Google doubles its AI updates impressively, and this time, we get better reasoning skills than ever.
Available now, you can access Gemini 2.5 PRO experimental simply by selecting the model in the Gemini application or directly in the AI Studio of Google. You will need an advanced Gemini subscription to see this as an option.
The prices of the improved model (for those who wish to use it for use of production on a scale) will be announced in the coming weeks, and more models 2.5 should be launched in due course.
Gemini 2.5 Pro is impressive
Google shared certain reference results for Gemini 2.5 pro experimental and the results are really impressive.
The new AI model marks 18.8% on the latest examination of humanity against 14% for the O3-Mini of Chatgpt and 8.6% for Deepseek R1. The last examination of humanity is the most in -depth and difficult AI reference, therefore marking considerably more than its competitors is not an easy task.
18.8% is the highest score that we have ever seen during the last examination of humanity (without the use of tools). Google calls for the reasoning capacities of Gemini 2.5 Pro “at the cutting edge of technology” and it is clear to see why.
Google continues to move forward with its development of AI at a rapid pace. Last week, the company made Gemini Deep Research for everyone and followed this with improvements in its impressive AI podcasting tool, Notebooklm.
We will test Gemini 2.5 PRO and put the new experimental model to the test, so stay listening to Techradar for more Google AI coverage.
