do son 
Following the announcement of Gemini 2.0 late last year, Google has now unveiled the further enhanced Gemini 2.5, introducing a fully-featured experimental version dubbed Gemini 2.5 Pro. The company highlights its remarkable capabilities in programming, mathematical reasoning, and scientific analysis.
Koray Kavukcuoglu, Chief Technology Officer of Google DeepMind, noted that Gemini 2.5 benefits from a significantly upgraded foundational model and refined post-training processes, resulting in a substantial leap in inference power. The model is now better equipped to tackle complex problems and serve as a more capable AI agent.
Unlike the earlier launch of Gemini 2.0, which debuted with the lighter Flash Thinking variant, this time Google has opted to introduce the full-fledged Gemini 2.5 Pro Experimental version from the outset—an evident move to remain competitive in a rapidly evolving AI landscape, particularly in response to emerging rivals such as China’s DeepSeek R1. Additional iterations of the model are expected to follow.
The Gemini 2.5 Pro Experimental model is initially available through Google AI Studio, and users subscribed to the Gemini Advanced service can also access it directly via the Gemini App.
